• TikTok is making its algorithm optional for users in the European Union to comply with the Digital Services Act.
• It agrees to laws that ban targeted advertisements for 13 to 17 year olds.
  

Next week, a law policing Big Tech in the European Union comes into force, starting the clock on a process expected to force companies like Amazon, Google, Meta, TikTok, and Twitter, among others, to make their platforms more open and interoperable. The Digital Services Act addresses social media’s societal harms by requiring companies to watch over their platforms for illicit content more aggressively or risk billions of dollars in fines.

READ NEXT

TikTok hit with multi-million dollar fine over child data safety

To recall, the European Parliament and all EU member states reached a political agreement on the Digital Services Act in late 2022. The legislation had set a timeline for companies under the Very Large Online Platforms (VLOPs) category to crack down on hate speech, disinformation, and other harmful and illegal material on their platforms. 

 “If the platform or a search engine has more than 45 million users (10% of the population in Europe), the Commission will designate the service as a very large online platform or a very large online search engine. These services will have four months to comply with the obligations of the DSA, which includes carrying out and providing the Commission with their first annual risk assessment,” the Commission stated.

Timeline for Digital Services Act as set by the European Commission.

On April 23, 2023, the European Commission named a list of 19 tech platforms required to comply starting August 25, 2023. The list includes 17 VLOPs and 2 Very Large Online Search Engines (VLOSEs).

• Alibaba AliExpress
• Amazon Store
• Apple AppStore
• Booking.com
• Facebook
• Google Play
• Google Maps
• Google Shopping
• Instagram
• LinkedIn
• Pinterest
• Snapchat
• TikTok
• Twitter
• Wikipedia
• YouTube
• Zalando

Meanwhile, the VLOSEs listed included Bing and Google Search. What comes after the designation is that the companies are required to comply, within four months, with the complete set of new obligations under the Digital Services Act. Overall, the new law is also helping make Brussels a trailblazer in the growing global movement to clamp down on tech giants. 

In light of the requirements by the European Commission, TikTok’s owner ByteDance and other large online platforms will be required to police illegal content on their platforms, prohibit certain advertising practises, and share data with authorities. TikTok quickly ensured they were on the good side of the block. 

On August 4, the video-sharing platform released new measures and features for European users to comply with the incoming Digital Services Act.

How will TikTok abide by the EU Digital Services Act?

Before releasing its new guidelines, TikTok agreed to a voluntary “stress test” last month. The result was that EU industry chief Thierry Breton recommended “more work” was needed for the firm to be fully compliant.

YOU MIGHT LIKE

TIKTOK

UK bans TikTok on government phones

“In the coming weeks, we are introducing an additional reporting option for our European* community that will allow people to report content they believe is illegal, including advertising,” TikTok said in a blog posting.

That means making this as easy as possible, and people can choose from categories such as hate speech, harassment, and financial crimes. “We will provide a guide to help people better understand each category,” TikTok noted. As for content reported as being illegal, it will first be reviewed against the platform’s ‘Community Guidelines’ and removed globally if it violates TikTok’s policies. 

If it does not, TikTok has a new dedicated team of moderators and legal specialists to assess whether it violates the law, and “we will restrict access to the content in that country only,” TikTok added. Under the DSA, it will inform European users about a broader range of content moderation decisions. 

“For example, if we decide a video is ineligible for a recommendation because it contains unverified claims about an election that is still unfolding, we will let users know. We will also share more detail about these decisions, including whether the action was taken by automated technology, and we will explain how both content creators and those who file a report can appeal a decision,” the blog post reads.

TikTok also allows its European users to turn off personalization for their feed to meet the Digital Services Act requirements. “This means a user’s For You and LIVE feeds will instead show popular videos from both the places where they live and around the world, rather than recommending content to them based on their interests.”

When using a non-personalized search, users will see results of popular content from their region and in their preferred language. Their Following and Friends feeds will continue to show creators they follow, but in chronological order rather than based on the viewer’s profile.

One of the most significant changes made to cater to TikTok’s European demographic is how it’ll protect teens’ privacy. Firstly, TikTok said accounts for those under 16 are private by default, and their content cannot be recommended in For You feeds. European users aged 13-17 will no longer see personalized advertising based on their activities on or off TikTok. “People already have control over the ads they can see, and they can toggle personalized ads on or off in their settings,” TikTok concluded.

The post TikTok bends to comply with the EU’s Digital Services Act appeared first on TechHQ.

Worldcoin, founded by Alex Blania and Sam Altman – who are both still involved – and Max Novendstern, who left the company in 2021, has some lofty ambitions. As CEO of OpenAI, Altman has brought advanced chatbots powered by generative AI into the mainstream. And Blania, Worldcoin’s CEO, brings an intriguing mix of theoretical physics and protocol development skills to the project, which sets its sights on creating the biggest financial and identity network imaginable. However, its popularity encountered a speedbump this week, with Worldcoin going down on Monday 7^th August.

But if Worldcoin being overwhelmed by demand is a measure of the global interest in the project, then Blania and Altman would have reason to view Worldcoin being down as another milestone in their journey. And the end goal is to solve the problem of verified humanness, which tackles the issue of identity in a world where telling the difference between human-generated and machine-made output is getting harder.

Why do we need a World ID?

Picturing a world where conversing with advanced chatbots – such as descendants of OpenAI’s game-changing ChatGPT – is indistinguishable from human interactions, having proof of personhood feels like a smart idea. In principle, interactions could then be labeled as human in origin to help people navigate what is likely to be a very confusing existence as generative AI becomes increasingly lifelike. But that raises the question, how do you verify a human?

Maxed out: Worldcoin was down on Monday 7th August, 2023, and returned minus the capability to claim grants as developers worked to restore services.

Worldcoin’s approach is based on liveness detection, biometrics, and so-called zero-knowledge proofs that allow systems to validate whether something is genuine while preserving the privacy of that information. And privacy-preserving validation of personhood opens the door – Blania and Altman hope – to some radical concepts such as universal basic income (UBI), as signposted in the Worldcoin whitepaper.

In fact, according to that whitepaper, every human is eligible for a share of Worldcoin simply for being human. But distributing those tokens fairly is only possible if users can verify their humanness – asserting that they are real people and different from other real people.

READ NEXT

Worldcoin and World ID – identity security against generative AI bots?

Conventionally – for example, when interacting with regular institutions – that happens using know-your-customer (KYC) processes such as providing a birth certificate or passport. But the founding team’s concern is that this approach doesn’t scale. There’s also the issue of what happens if you don’t have those documents.

Considering alternative approaches, Worldcoin has opted for biometrics – specifically, iris recognition. And having a digital primitive based on the pattern of a user’s eye opens up the concept to a truly global audience.

Recall that Apple’s Face ID device unlocking feature carries a probability that one person in a million, other than you, could look at the infrared camera and gain access to the smartphone or tablet. And, although that probability sounds small – if you wanted to use that method as a proof of personhood on a global scale for billions of Worldcoin accounts, the risk of fraud would be too high.

A bridge to the real world

Hearing Blania and Altman speak, it’s clear that they can imagine a future where looking at pictures or video, or considering intelligence alone, won’t be sufficient to determine humanness. And they see World ID’s proof of personhood as a bridge to the real world, which is a good time to mention the Worldcoin Orb.

Shoreditch sign-up: it takes just a few minutes for users to download the Worldcoin app, agree to the T&C’s, and verify that they are human by standing in front of a Tools for Humanity Orb, which takes an infrared measurement to determine liveness and performs an iris scan. Image credit: JT.

Last week, when Worldcoin wasn’t down, this author gazed briefly into one of 2000 Orbs that have been produced by Tools For Humanity – a technology company building tools for the Worldcoin project. Custom designed, Worldcoin’s Orbs feature a range of sensors, an iris scanner, and contain local compute that turns images of the colored part of the eye surrounding the pupil into device-signed World IDs.

By presenting a digital World ID, users can assert that they are a unique human who has stood in front of a Worldcoin Orb, and for signing up they receive – depending on where they reside – a nominal amount of Worldcoin tokens. And this goes back to the idea that by offering rewards, users will be incentivized to make the network a success.

However, not everyone is so keen on the idea of a new global currency – notably, the US – and restrictions are in place that stop Worldcoin from issuing tokens to US residents. Also, in Europe, governments appear to share some concerns about the collection of sensitive information.

READ NEXT

OpenAI quietly bins its AI detection tool – because it doesn’t work

For example, the Information Commissioner’s Office (ICO) in the UK advises that “Organisations must conduct a Data Protection Impact Assessment (DPIA) before starting any processing that is likely to result in high risk, such as processing special category biometric data.” And, “Where they identify high risks that they cannot mitigate, they must consult the ICO.”

However, while the ICO has said in a statement on Worldcoin that it will be making enquiries, the independent body has dodged the question – based on the details of a freedom of information request – as to whether it has asked Worldcoin for a copy of its DPIA.

Worldcoin is processing biometric data, but the processing – the conversion of the iris scan into a unique World ID – takes place locally on the Orb via a neural network embedding. And it’s interesting to note that the hardware, while custom-designed to capture sufficiently high-resolution images capable of preserving uniqueness on a global scale, has been engineered to be resistant to physical attacks.

Blania, who has knowledge of how the Orbs have been engineered, revealed that the units make use of multi-spectral imaging and time-of-flight measurements to confirm that they are looking at real humans. And when proof of personhood takes place, it’s using an iris code that he describes as providing extreme privacy.

The zero-knowledge proofs in the protocol are intended to separate a user’s identity from any requests made to the World App for human verification.

How do I sign up for a World ID?

Practically, users need to book an appointment where they can visit a Worldcoin Orb near them and download the World App on their smartphone. The World App generates a QR code, which – in effect – says that the user is (potentially) the rightful owner of the World ID. And showing the QR code to the Worldcoin Orb initiates the final stage of the enrolment process – proving the person signing up is human.

Under the hood is the generation of key pairs – one pair for the World ID and another for a cryptocurrency wallet. And the zero-knowledge proof rests on whether the digital codes are part of a known set – in other words, proof of inclusion.

Considering use cases, solutions like World ID – and it should be said that there are other projects too, developing proof of personhood tools (Quadrata, Humanbound, and Proof of Humanity are just a few examples) – pave the way for authors to identify content as human-created. And others could use their proof of personhood to endorse human-badged creations.

Real deal: Orb verified users receive a proof of personhood, which can be viewed digitally on the World App and is linked to their World ID.

The Worldcoin project has attempted to design the protocol so that it can be scaled to all of humanity. However, until that backend becomes decentralized – for example, to piggyback on the benefits of federated computing – there will no doubt be growing pains to keep pace with interest in the project. And that’s been seen this week with Worldcoin down, and having to throttle functionality.

At the time of writing, the World App remains at capacity, according to a warning notification displayed on the World ID page of this author’s device.

The post Worldcoin down but not out appeared first on TechHQ.

If somebody told you that a refurbished laptop could eclipse the performance of an NVIDIA A100 GPU when training a 200 million-parameter neural network, you’d want to know the secret. Running AI routines on CPUs is supposed to be slow, which is why GPUs are in high demand, and NVIDIA shareholders are celebrating. But maybe it’s not that simple.

Part of the issue is that the development and availability of GPUs, which can massively parallelize matrix multiplications, has made it possible to brute force progress in AI. Bigger is better when it comes to both the amount of data used to train neural networks and the size of the models, reflected in the number of parameters.

READ NEXT

Can zero-trust LLMs overcome poisoned generative AI?

Considering state-of-the-art large language models (LLMs) such as OpenAI’s GPT-4, the number of parameters is now measured in the billions. And training what is, in effect, a vast, multi-layered equation – by first specifying model weights at random and then refining those parameters through backpropagation and gradient descent – is now firmly GPU territory.

Nobody runs high-performance AI routines on CPUs, or at least that’s the majority view. The growth in model size, driven by the gains in accuracy, has led users to overwhelmingly favor much faster GPUs to carry out billions of calculations back and forth.

But the scale of the latest generative AI models is putting this brute force GPU approach to the test. And many developers no longer have the time, money, or computing resources to compete – fine-tuning billions of artificial neurons that comprise the many-layered networks.

ThirdAI's new PocketLLM app is free to use and completely secure. https://t.co/e9uEfj1LmZ

— InnovationMap (@InnoMapHou) May 16, 2023

Experts in the field are asking if there’s another, more efficient way of training neural networks to perform tasks such as image recognition, product recommendation, and natural language processing (NLP) search.

Artificial neural networks are compared to the workings of the human brain. But the comparison is a loose one as the human brain operates using the power of a dim light bulb, whereas state-of-the-art AI models require vast amounts of power, have worryingly large carbon footprints, and require large amounts of cooling.

That being said, the human brain consumes a considerable amount of energy compared with other organs in the body. But its orders of magnitude GPU-beating capabilities stem from the fact that the brain’s chemistry only recruits the neurons that it needs – rather than having to perform calculations in bulk.

READ NEXT

What can generative AI do for business?

AI developers are trying to mimic those brain-like efficiencies in computing hardware by engineering architectures known as spiking neural networks. Neurons behave more like accumulators and fire only when repeatedly prompted. But it’s a work in progress.

However, it’s long been known that training AI algorithms could be made much more efficient. Matrix multiplications assume dense computations, but researchers have shown a decade ago that just picking the top ten percent of neuron activations will still produce high-quality results.

The issue is that to identify the top ten percent you would still have to run all of those sums in bulk, which would remain wasteful. But what if you could look up a list of those most active neurons based on a given input?

And it’s the answer to this question that opens up the path to running AI on CPUs, which is potentially game-changing – as the observation that a refurbished laptop can eclipse the performance of an NVIDIA A100 GPU hints at.

How to run AI on CPUs

So what is this magic? At the heart of the approach is the use of hash tables, which famously run in constant time (or thereabouts). In other words, searching for an entry in a hash table is independent of the number of locations. And Google puts this principle to work on its web search.

For example, if you type ‘Best restaurants in London’ into Google Chrome, that query – thanks to hashing, which turns the input into a unique fingerprint – provides the index to a list of topical websites that Google has filed away at that location. And it’s why, despite having billions of websites stored in its vast index, Google can deliver search results to users in a matter of milliseconds.

And, just as your search query – in effect – provides a lookup address for Google, a similar approach can be used to identify which artificial neurons are most strongly associated with a piece of training data, such as a picture of a cat.

In neural networks, hash tables can be used to tell the algorithm which activations need to be calculated, dramatically reducing the computational burden to a fraction of brute force methods, which makes it possible to run AI on CPUs.

In fact, the class of hash functions that turn out to be most useful are dubbed locally sensitive hash (LSH) functions. Regular hash functions are great for fast memory addressing and duplicate detection, whereas locally sensitive hash functions provide near-duplicate detection.

Dynamic sparsity

LSH functions can be used to hash data points that are near to each other – in other words, similar – into the same buckets with high probability. And this, in terms of deep learning, dramatically improves the sampling performance during model training.

Hash functions can also be used to improve the user experience once models have been trained. And computer scientists based in the US at Rice University, Texas, Stanford University, California, and from the Pocket LLM pioneer ThirdAI, have proposed a method dubbed HALOS: Hashing Large Output Space for Cheap Inference, which speeds up the process without compromising model performance.

As the team explains, HALOS reduces inference into sub-linear computation by selectively activating only a small set of likely-to-be-relevant output layer neurons. “Given a query vector, the computation can be focused on a tiny subset of the large database,” write the authors in their conference paper. “Our extensive evaluations show that HALOS matches or even outperforms the accuracy of given models with 21× speed up and 87% energy reduction.”

Field test

Commercially, this approach is helping merchants such as Wayfair – an online retailer that enables customers to find millions of products for their homes. Over the years, the firm has worked hard to improve its recommendation engine, noting a study by Amazon that even a 100-millisecond delay in serving results can put a noticeable dent in sales.

And, sticking briefly with online shopping habits, more recent findings published by Akamai report that over half of mobile website visitors will leave a page that takes more than three seconds to load – food for thought as half of consumers are said to browse for products and services on their smartphones.

All of this puts pressure on claims that clever use of hash functions can enable AI to run on CPUs. But the approach more than lived up to expectations, as Wayfair has confirmed in a blog post. “We were able to train our version three classifier model on commodity CPUs, while at the same time achieve a markedly lower latency rate,” commented Weiyi Sun – Associate Director of Machine Learning at the company.

Plus, as the computer scientists described in their study, the use of hash-based processing algorithms accelerated inference too.

The post Is running AI on CPUs making a comeback? appeared first on TechHQ.

• The new logo of Twitter marks a new era for the social media platform that has been known for its blue bird and ‘tweets’ for over 17 years.
• Analysts view the changes as a way for Musk to achieve his vision that the app could offer various services to users beyond social media.

It is time to savor the final days of Twitter in its original state — the social media platform is at the brink of its most drastic transformation, and it’s more than just a brand new logo. There is no denying that Twitter has undergone drastic changes over the last eight months since Elon Musk completed his US$44 billion deal to own the platform. 

READ NEXT

The social media rulebook: content moderation

But none have come close to what the 17-year-old social platform is enduring this week. Its iconic blue bird is spending its last hours as the app’s symbol. Started in the US on July 24 this week, the app and its website are gradually being transformed to be known as ‘X,’ and tweets will be dubbed ‘x’s.’

It all started on July 22 when Elon Musk tweeted, “Soon we shall bid adieu to the Twitter brand and, gradually, all the birds.” The next 48 hours were followed by speculations on what the rebranding could mean and what new logo would replace the iconic blue bird. On the afternoon of July 23, Sunday, Musk tweeted an indication that “If a good enough X logo is posted tonight, we’ll go live worldwide tomorrow.” 

Source: Twitter

Within the next 30 minutes, a user that goes by the name @SawyerMerritt posted a three seconds video of what would potentially be the new Twitter logo. Musk retweeted the video, indicating his preference for that particularly fan-made design. Not too long after, Musk changed his profile image to the new logo. He even posted a picture of the structure projected on Twitter’s San Francisco headquarters.

The transition happened too quickly for many to grasp. By the morning of July 24, Twitter began replacing its decade-long logo with a fan-made “𝕏” logo. It was all over Twitter’s homepage, as a profile picture for its official Twitter account, and on a splash screen displayed while the website loads. However, it is worth noting that the blue bird logo hasn’t been erased from the service entirely — it still serves as the website’s favicon and remains prominent throughout the mobile apps. 

What does the new logo mean for Twitter?

Looking back, the rebrand comes after months of erratic behavior by Musk, which impacted users and pushed away advertisers, leaving Twitter in a troubled financial position and increasingly vulnerable to competition. For context, Musk had already converted Twitter’s corporate name to X Corp, a subsidiary of X Holding Corp, as revealed in an April court filing. 

Musk has used the letter X repeatedly across his companies. He co-founded x.com as an online bank in 1999, which later transformed into PayPal. He repurchased the domain from PayPal in 2017, saying it had “sentimental value.” In the US, the domain x.com now redirects to Twitter. 

X.com now points to twitter.com. Source: Twitter

“The new logo garnered mixed reactions from users and sparked confusion about what tweets would now be called, while marketing and branding experts said the rebrand risked throwing away years of Twitter’s name recognition,” Reuters stated in an article yesterday. 

Source: Elon Musk’s Twitter

Before buying Twitter, Musk had even said last October that he viewed the US$44 billion deal as “an accelerant to creating X, the everything app.” In fact, for as long as Twitter has been on Musk’s mind, he has envisioned an app that could offer various services to users beyond social media, such as peer-to-peer payments. The idea mirrors the widely popular WeChat app in China. 

Quoting Tom Morton, global chief strategy officer at ad agency R/GA, Reuters stated that the transformation is simply a way for Musk to make his mark on the company. “Twitter’s changing name and logo has nothing to do with user, advertiser, or market issues. It symbolizes that Twitter is Elon Musk’s personal property,” Morton told Reuters.

READ NEXT

Why (and How) is Social Media Failing Businesses?

Linda Yaccarino, Twitter’s CEO since June 5, also told employees in a memo on Monday that X “will go even further to transform the global town square.” And its believed that the company will work on new audio, video, messaging, payments, and banking features.

For now, the move of Musk’s renaming of Twitter as X is still fresh, and it can either turn out to be one of his biggest missteps since buying the company or a stroke of brilliance. There are arguments on both sides. But, indeed, a sharp name change and a new logo could help change perceptions about a tattered brand like Twitter. 

The social wind of change?

The wind of change has not only been apparent in Twitter; there has been a surge of activity in the social media space over the last few weeks. Besides Twitter’s rebrand and the launch of Threads to take on Twitter’s microblogging dominance, TikTok has announced a new text feature this week that will broaden the kinds of content that creators can share with followers.

The short-form video app allows sharing text-based posts with music and stickers, similar to Instagram Stories, and we’ll be digging into the impact of the update once the new TikTok feature enters the mainstream market.

Creators have been able to make content on TikTok across a variety of formats — from LIVE videos to photos, Duets to Stitch. Text is the latest addition to options for content creation, allowing creators to share their stories, poems, recipes, and other written content on TikTok – giving creators another way to express themselves and making it even easier to create.
Source: TikTok

The post It is the end of the blue bird era for Twitter. Here comes Musk’s ‘X’ appeared first on TechHQ.

• Apple needs a clear strategy for releasing AI technology to consumers.
• Rumors that Apple has been left behind in AI look set to be confounded next year.
• Apple has been pushing forward with AI in small ways.

Since OpenAI’s ChatGPT debuted last November, tech leaders have not been very subtle about the potential of AI and their plans for the technology. Apple Inc. is the only big tech firm that has yet to jump on the generative AI bandwagon. But, like the rest of Silicon Valley, the company may soon take bigger swings.

READ NEXT

Meta, Microsoft release new AI language model for commercial use

All this while, Apple has been approaching AI cautiously, with a product-oriented focus. Even during Apple’s annual developer’s conference this year, the WWDC 2023, the company subtly touted how much work it was doing in state-of-the-art AI and machine learning.

With its usual determination to be different, Apple doesn’t actually use the term “artificial intelligence” or even the more precise “generative AI” — it prefers the more academic phrase “machine learning” or talks about the features the technology enables. One example of this strategic coyness is that during the WWDC 2023, Apple announced that AirPods will now use “Adaptive Audio” to analyze sound around its user and adjust themselves accordingly. The feature allows the AirPods Pro to automatically turn off noise canceling when the user engages in conversation.

Although Apple didn’t frame that as a machine learning feature, the solution is based on AI models. On top of that, the company has made AI headway in other areas, including improvements to Photos and iPhone searches. There’s also a smarter version of auto-correct coming to its mobile devices this year.

Apple also mentioned several other new features that used the company’s skill in neural networks, such as the ability to identify fields to fill out in a PDF. The Photos app will also be able to differentiate between your dog and other dogs, automatically recognizing your pup the same way it identifies people who frequently appear in your pictures. 

A technology analyst from Wedbush Securities, Daniel Ives, told The Atlantic that he thinks Apple’s new AI features amount to “just the appetizer before the main entrée.” His team has estimated that the company has spent US$8 to US$10 billion on AI in the past four or five years—the same amount that Microsoft invested in OpenAI in January—and Apple is reportedly on the hunt for AI talent.

However, its main AI product, the Siri voice assistant, has stagnated in recent years. But from how Chief Executive Officer Tim Cook put it during a conference call in May, Apple will be adding AI to more of its products, but, he said, on a “very thoughtful basis.”

Cook also said that though the technology has potential, a “number of issues still need to be sorted.” However, unlike its rivals, Apple wants AI models on its devices, which build bigger models with server farms, supercomputers, and terabytes of data. 

“The new autocorrect feature is imposing because it’s running on the iPhone, while models like ChatGPT require hundreds of expensive GPUs working in tandem,” a report by CNBC indicated.

Apple GPT – an AI chatbot in iPhones soon?

A consumer “Apple AI” product is aimed for next year. Source: Twitter

Apple has been so quiet about its AI progress that it has been accused of falling behind. While ChatGPT can write you an essay, Siri can set your morning alarm and not much else. But hints indicate Apple is pushing forward with AI in small ways.

Albeit advancing by the incrementalist approach, it nonetheless still might be the future of where this technology is headed. Indeed, according to people who know the efforts, the iPhone maker has built its framework to create large language models — the AI-based systems at the heart of new offerings like ChatGPT and Google’s Bard.

“With that foundation, known as ‘Ajax,’ Apple also has created a chatbot service that some engineers call ‘Apple GPT,'” according to a report by Bloomberg. Perhaps, behind the scenes, Apple has grown concerned about missing a potentially paramount shift in how devices operate. “And Apple’s devices, which produced revenue of nearly US$320 billion in the last fiscal year, could suffer if the company doesn’t keep up with AI advances,” the report added.

YOU MIGHT LIKE

BIG DATA

Can zero-trust LLMs overcome poisoned generative AI?

Apparently, in recent months, the AI push has become a significant effort for Apple, with several teams collaborating on the project, sources told Bloomberg. The work includes trying to address potential privacy concerns related to the technology. For context, on-device AI bypasses many data privacy issues that cloud-based AI faces, and Apple needs to collect less data when the model can be run on a phone.

There may be fewer privacy issues on an Apple AI running on-device.

It also ties in closely with Apple’s control of its hardware stack, down to its silicon chips. Apple packs new AI circuits and GPUs into its chips every year, and its control of the overall architecture allows it to adapt to changes and new techniques. “That’s why Apple began laying the foundation for AI services with the Ajax framework, as well as a ChatGPT-like tool for use internally,” Bloomberg said.

What exactly is the status of “Apple GPT”?

Simply put, Apple is developing new AI tools but still needs to figure out how to commercialize them. Ajax was first created last year to unify machine learning development at Apple, according to the people familiar with the effort. Then a tiny engineering team made the chatbot app as an experiment at the end of last year. 

If you’re Apple, how do you commercialize your investment in AI?

“Its rollout within Apple was initially halted over security concerns about generative AI, but has since been extended to more employees. Still, the system requires special approval for access. There’s also a significant caveat: any output from it can’t be used to develop features bound for customers,” Bloomberg said, according to sources.

READ NEXT

Apple to lead the AR/VR headset market into the mainstream

Apple employees indicated the tool replicates Bard, ChatGPT, and Bing AI and doesn’t include novel features or technology. The system is accessible as a web application and has a stripped-down design not meant for public consumption. “As such, Apple has no current plans to release it to consumers, though it is actively working to improve its underlying models,” Bloomberg said.

The conclusion, for now, lies in the fact that beyond the state of the technology, Apple is still trying to determine the consumer angle for generative AI. While the company doesn’t yet have a concrete plan, Bloomberg said people familiar with the work believe Apple aims to make a significant AI-related announcement next year.

The post Apple GPT: An AI work-in-progress? appeared first on TechHQ.

We are living in boom times for GPUs. The huge appetite for augmenting software with large language models (LLMs) and building advanced chatbots is great news for chip designers such as NVIDIA, whose hardware powers generative AI. Users are scrambling for LLM training stacks and inference architecture, leading to supply issues. But what if you could sidestep GPU shortages entirely by running deep learning algorithms on CPUs instead?

READ NEXT

Airbnb for GPUs – decentralized IaaS for generative AI

It’s a radical idea that some analysts believe could impact NVIDIA’s stock price in the long term. And not only would switching to currently underutilized CPUs address the problem of GPU shortages, energy-efficient approaches to training LLMs on CPUs bring other benefits to the table as well.

Fed with huge datasets, LLMs can take months to train and optimize, consuming large amounts of energy and water (generative AI is thirsty for cooling) in the process. And environmentalists are concerned about the carbon footprint of having banks of GPUs running 24/7 as developers race to build ever more powerful LLMs and push ahead of the competition.

Dynamic sparsity saves resources

OpenAI’s GPT-3, which provided the foundation for the hugely successful ChatGPT, has 175 billion parameters – the various weights that determine how deep learning neural networks map inputs into outputs. And, as it turns out, this number can be made dynamically much smaller without impacting model accuracy, which is a game changer for democratizing AI.

Being able to train deep learning models with hundreds of thousands of input dimensions and several thousand hidden layers on a CPU makes generative AI much more portable and standalone. So-called pocket LLMs – which could be carried in a backpack and kept offline – provide air-gapped generative AI capabilities that are ideal for keeping company data safe.

In the excitement to see what’s possible using conversational AI, many users may have inadvertently submitted sensitive data to LLMs hosted in the cloud. AI heavyweights such as Open AI and Google warn users not to enter sensitive information – pointing out that human reviewers may process conversations for quality purposes.

Big players have invested substantial sums in reinforcement learning with human feedback to make advanced chatbots the success story that they’ve become. And air-gapped, pocket LLMs running on CPUs give users the chance to bring that model refinement in-house and take full control over the development of domain-specific solutions trained with private data.

Chips in high demand – why are there GPU shortages?

GPUs became all the rage in the AI community when developers saw that bigger was better. Large models with many hidden layers fed with huge amounts of data produced rubbish to begin with. But leaving that compute whirring away for days – back-propagating results to optimize the weights of the neural network, and completing multiple passes (epochs) through the whole training set – produced staggering results.

In the case of next-word predicting LLMs, data doesn’t even have to be labelled. Models are shown sentences with words removed and have to guess what’s missing. Incredibly, when performed at scale, this unsupervised learning method is capable of teaching computers how to translate languages, write code, and converse convincingly with humans – to list just a few wonders of generative AI.

READ NEXT

Business intelligence enters the age of the smart summary

Under the hood, words are broken down into sub-parts known as tokens and stored in large vectors that are multiplied together. And it’s the ability to efficiently perform vector multiplication that gives GPUs the edge, hence the high demand for chips from leading designers such as NVIDIA.

GPUs have relatively small instruction sets, but large numbers of cores. Originally, they were devised to perform simple tasks – turning pixels on and off – but rapidly, in parallel at scale. And today, GPU use goes beyond graphics, having proven to be popular when mining cryptocurrency was all the rage, and more recently unpinning the latest AI boom.

But for how long? As we hinted at earlier, dynamic sparsity changes things by only considering the AI model parameters that are necessary for the current sample. Training deep learning algorithms on GPUs is a brute-force approach that performs billions of vector multiplications whether they are needed or not. In reality, only a few thousand model parameters may actually be updated when shown a new input, with the bulk of the calculations simply multiplying something by zero.

Bolt dynamic sparsity engine for CPUs

By focusing the compute on the high activations and ignoring the low activations, developers find that it’s possible to use commonly available CPUs – even though such chips have fewer cores. And users can once again benefit from the superior memory capacity of CPUs – one of the constraints of fast, parallel-processing GPUs.

“People haven’t seen what CPUs can do,” Anshu Shrivastava CEO and Founder of ThirdAI (pronounced third-eye) told TechHQ. “We’ve done all the heavy lifting to make model training very efficient.”

Rather than having to pipe their business intelligence into a third-party service, users can instead – thanks to ThirdAI’s dynamic sparsity engine, dubbed Bolt – build and deploy billion parameter models that run on their own CPUs. And having that information locally, in-house, makes it so much easier to keep those generative AI-solutions – which could be internal company chatbots, or highly interactive document search tools – up to date.

“Every data point that goes through infrerence can become part of a new training set,” adds Shrivastava. “It’s a continuous process.”

The post Like the sound of private, air-gapped LLMs? Welcome to generative AI on CPUs. appeared first on TechHQ.

DuckDuckGo’s Windows browser went into public beta last month, and TechHQ has been putting the privacy-focused search engine to the test – to discover the upsides and downsides of switching from dominant players such as Google Chrome.

Installing DuckDuckGo on Windows

Firstly, let’s start with the installation process. Windows users simply download the app installer, click install, and DuckDuckGo is loaded in the time it takes to drink a coffee.

Once installed, the browser launches and gives you the option to import your bookmarks and passwords (if you have them saved to your current internet search engine).

READ NEXT

AI in search – Neeva closes answer engine, switches to enterprise

Users can then choose to set DuckDuckGo as their default browser. But let’s see what the software can do first, before making DuckDuckGo our our first pick for internet search. And completing the app walkthrough, the last of the first-time guide windows tells us that we’re all set.

DuckDuckGo badges itself as a ‘privacy browser’ and the search tool aims to ‘block trackers and upgrade the security of your connection when possible’.

Benefits of using DuckDuckGo

DuckDuckGo’s tagline is ‘Privacy simplified. All for free.’ And the full feature set can be viewed by clicking on the ‘Get Browsing Protection’ button, which promises the following –

• Search privately
• Browse more securely
• Browse more privately
• Instantly erase browsing activity
• Block annoying cookie pop-ups
• Block email trackers
• Block app trackers for your android device
• Watch Youtube with fewer distractions & more privacy
• See who’s trying to track you

Microsoft Edge Webview2?

Clicking the ‘About DuckDuckGo’ button shows the version number and gives details of the technology under the hood in the Windows app. The privacy-focused search engine uses Microsoft Edge Webview2, and DuckDuckGo is quick to point out that its engineers have spent lots of time addressing privacy issues specific to WebView2 – for example, by ensuring that crash reports are not sent to Microsoft.

Peter Dolanjski – Product Director at DuckDuckGo – emphasizes that their Windows app is not a fork of any other browser code. “All the code, from tab and bookmark management to our new tab page to our password manager, is written by our own engineers,” he explains.

DuckDuckGo’s UI team has done a great job with the ‘All Settings’ page, breaking down all of the user preferences across three easy-to-digest tabs – General, Appearance, and Privacy. And there’s some fun to be had here with a bit of color hex code research.

Color customization win

If you are a fan of the Chicago Bulls, Paris Saint Germain, Chiefs Esports Club, the Mercedes AMG Petronas F1 Team, or any sports outfit, then you’ll enjoy DuckDuckGo’s color customization options.

Go faster stripes: the DuckDuckGo browser has a high-level of color customization, which users can have fun with.

The web browser allows users to dive in and specify hex codes for the background color, header color, title color, visited title color, snippet color, URL color, hover, module, and dropdown background color. And thanks to resources such as Team Color Codes, modding DuckDuckGo to render in your favourite team kit requires just a few clicks.

Is this feature essential? No. Is it fun to mod your web browser so that it’s on-brand with the Mercedes AMG Petronas F1 team? Yes.

Busy behind the scenes

And bringing things back on topic, it’s interesting to note that – according to DuckDuckGo’s reporting stats, which appear below the search bar – the privacy-focused browser has been busy. While this reviewer was scrolling through teamcolorcodes.com, DuckDuckGo reportedly blocked 33 tracking attempts (which has climbed to 42 tracking attempts while writing the rest of this story) made by the website.

Choosing to batten down the hatches on privacy may come with some inconvenience if you do appreciate being remembered by the websites that you visit. But one of the advantages for users of DuckDuckGo is that they can start with privacy defenses set high, and then choose to lower them once they have visibility on who’s attempting to track them and why.

Other ecosystems available: DuckDuckGo supports Mac, iPhone, and Android systems as well as Windows. Image credit: DuckDuckGo.

NPR has an extremely informative and brilliantly titled podcast episode – How the cookie became a monster – that celebrates 30 years of a piece of code designed originally to solve a fundamental usability problem of the internet. Without cookies, online shopping carts would have no memory of which goods had been put in the basket, and e-commerce would be incredibly tedious.

But, having paved the way for a better online shopping experience, web cookies were found to have other talents, helping companies to better target advertising to users, which browser developers such as DuckDuckGo have pushed back on.

The problem of web-based device fingerprinting

And cookies aren’t the only way of tracking users browsing websites on the internet – there’s device fingerprinting too, which has been around for many years. “What if today I could tell you that interested parties could track users without the need of cookies, or any other stateful client-side identifiers?” security researcher Nick Nikiforakis, now at Stony Brook University, US, told an audience of web users at Microsoft Research in 2016.

Unlike cookies, users can’t delete properties of their machine, like device screen size and CPU type. Browser type and version, fonts, plug-ins, and other details – even blocked cookie requests – add up to a unique fingerprint of who we are the web. And – you’ve guessed it – these device fingerprints turn out to be a good proxy for all of those cookies and web trackers that privacy-focused users are spending their efforts blocking.

READ NEXT

Can blockchain-based search engines beat Google?

A landmark 2010 paper by Peter Eckersley of the Electronic Frontier Foundation (EFF) asked ‘How Unique Is Your Web Browser?’ [opens as PDF]. And if you are worried about privacy on the web, you may want to look away now as we recall the results. Based on device fingerprints collected from 470,161 browsers, during sessions initiated by volunteers who were aware of the nature of the experiment (with some even actively trying to obfuscate their behavior), 94.2% were uniquely identifiable, and 4.8% had fingerprints that were seen just twice.

If you’re thinking that it’s hard for regular folks to stay hidden on the web, you’d be right. What’s more, if you are in the minority in trying to stay hidden – for example, by using DuckDuckGo rather than Chrome, and running a security and privacy-enhanced operating system, such as GrapheneOS (if you have a Pixel phone) you could be increasing the number of tells, simply by appearing so different to the majority of other web users.

Whether you’d want to install Windows on a Dell laptop and run Google Chrome to blend in, is another question. But evading device fingerprinting entirely isn’t easy, and maybe the solution is to work with it and sow seeds of confusion.

DuckDuckGo blocks many fingerprinting scripts before they have a chance to load, but the browser has other tricks up its sleeve. “We override many of the browser APIs used for fingerprinting to make them return either no information or alternative information that’s less useful for fingerprinting,” write the developers.

Should you use DuckDuckGo as your main browser?

DuckDuckGo positions itself as a privacy-focused web browser. And it’s clear that the team takes this mission seriously. In 2022, DuckDuckGo’s annual donation program awarded a total of $1,100,000 to 24 privacy groups and organizations advocating internet user rights around the world. Beneficiaries include the EFF, the European Digital Rights network, Signal Technology Foundation, and many other names notable for defending the rights of web users to online safety and security.

But the choice of which web browser to use remains an individual one. And usability goes hand-in-hand with the app experience, especially if you spend a large amount of time online and regularly search the web as part of your job.

DuckDuckGo is looking for a senior C# developer to build cool stuff!#dotnet #csharp

Check out this job at DuckDuckGo: Senior Software Engineer, Windows Desktop App (Remote) https://t.co/D1l9lRSEX3

— Calvin A. Allen | Microsoft MVP (@_CalvinAllen) March 3, 2023

DuckDuckGo’s Windows beta is fairly easy to jump into if you’ve been using Chrome or Brave. And the search engine performs well in delivering useful results, closer to Chrome than Brave, based on this author’s experience.

DuckDuckGo is a so-called metasearch engine, gathering results through a variety of vendor APIs – aggregating the search capabilities of many different engines.

READ NEXT

ChatGPT versus Google – the future of search

Concluding that using a privacy-focused browser offers full protection to users as they search the internet feels misleading, given the sophistication of modern device fingerprinting. But what DuckDuckGo does do well, is put privacy center stage and gets users thinking about the topic in more detail.

For example, being able to quickly click on the shield in the search bar to see requests blocked from loading, allows users to become familiar with the top third-party trackers. And it brings the discussion around to why our internet search behaviour needs to be tracked to such a high degree.

DuckDuckGo is open about its business model, which includes private ads displayed alongside search results, with a difference. “Search ads on DuckDuckGo are based on the search results page you’re viewing instead of being based on you as a person,” writes DuckDuckGo’s commercial team. “For example, if you search for cars, we’ll show you ads about cars.”

Cautionary tale

There are buttons to report ads that seem suspicious or irrelevant, and there are cases where bad actors have used lesser-known search engines to try and fool users into clicking on fake websites. The blockchain-based search engine Kin was conceived after an attempted phishing attack was discovered in an advert presented by the search engine DuckDuckGo.

But again, progress is made by raising awareness. And a newer feature on DuckDuckGo is email protection – a free email forwarding service that removes multiple types of hidden email trackers and lets users create unlimited unique private email addresses on the fly. In beta-testing, the developers found that 85% of emails being handled by the service contained hidden trackers used to build a profile on message recipients.

And that’s the last word – DuckDuckGo’s big selling point is getting people talking about privacy when browsing the web and receiving emails.

The post Reviewing the new DuckDuckGo browser on Windows appeared first on TechHQ.

• Many looking for Twitter alternatives again.
• Read limits blamed on web scraping.
• Hive regarded as one of the better Twitter alternatives.

In its Elon Musk era, Twitter is different in almost every way from what it was before. Since the chief of Tesla and SpaceX acquired Twitter around a year ago, it has been quite a ride for the company and users alike. So much so that under the ownership of Musk, there have been one too many instances where Twitter users were seen flocking away, looking for alternatives.

READ NEXT

Twitter in transphobia storm amid executive exits

The most recent spectacle started on June 30, when Musk claimed that “several hundred organizations” were taking Twitter’s data in a process called scraping, and proclaimed that “it was affecting the real user experience.” So to curb the effects of data scraping, Musk decided that Twitter would no longer allow individuals to view tweets on the social media platform unless they logged into an account.

But that’s not all.

A day after Musk’s claim, thousands of users reported widespread problems in accessing Twitter. They received error messages stating they had “exceeded” their “rate limit,” violated Twitter’s rules, and downloaded/viewed too many tweets. The issue was soon followed by a new set of Musk restrictions.

He took to his Twitter account to announce that the platform would temporarily limit the daily posts users read to address concerns over data scraping. At first, without stating how long the limits would last or what could prompt him to lift the restriction, Musk shared that verified accounts were temporarily limited to reading 6,000 daily posts.

He further explained that unverified and new unverified accounts would be limited to reading 600 posts a day and 300 posts a day, respectively. Two hours later, the temporary limitation was then increased to 10,000 posts per day for verified users, 1,000 posts per day for unverified users, and 500 posts per day for new, unverified users.

Many Twitter users are reporting being hit with a “Rate Limit Exceeded” message while browsing or trying to tweet. Source: Twitter

Before the move, Musk was vocal about his dislike of organizations scraping Twitter for research or to train artificial intelligence programs. But limiting the number of tweets users can read was something no-one had anticipated. As with other frenetic Musk-era changes at Twitter, this one has motivated some users to try other alternatives.

Not the first search for Twitter alternatives

As TechCrunch puts it, interest in Twitter alternatives began in late October 2022, when the Tesla and SpaceX chief officially closed on his US$44 billion Twitter acquisition. “Continual chaos on the microblogging site has been the rule ever since. 

Amid controversial policy decisions and overnight changes, some subset of the Twitter audience decided to exit to other sites to get their social fix — or at least experiment with different options for a time,” the article aptly says.

YOU MIGHT LIKE

TWITTER

Twitter API vulnerability leaves millions exposed

Considering many remaining Twitter users once again found themselves at the same crossroads, we decided to look into the microblogging site alternatives that have been trending since the weekend. 

Alternatives benefiting from the latest Twitter exodus

While many alternatives are available today, here are three that most likely may pose a challenge to Twitter.

Mastodon

When Musk bought Twitter nearly six months ago, many users felt it was the right time to leave the platform. Thousands of tweeters fled to Mastodon: a social media project designed from its start in 2016 to resist takeovers by billionaires.

Among all the alternatives to Twitter, Mastodon is often cited as the most likely option. Source: Shutterstock

Independently run servers – each with its moderators and users – can interact with each other’s posts, called “toots,” using an open protocol called ActivityPub.

Even other social media services can connect to ActivityPub, so no one app can monopolize the broader network that Mastodon is part of, called the “diverse.” Users must also choose a server to determine who would charge their data and what toots they see most often.

Mastodon gained roughly 500,000 users within ten days of Musk’s Twitter takeover on October 27, 2022. By the end of November 2022, there were 2.5 million users registered to the online network, an increase of around 300% within five months. 

According to Statista, as of March 2023, the decentralized social media platform had over ten million registered users. There is, however, no clear indication of the number of users Mastodon has gained following Twitter’s latest havoc.

Bluesky

Still an invite-only app in its beta, Bluesky has around 50,000 users, and according to estimates from data.ai, the app has been downloaded more than 375,000 times. Which means many people are trying to get an invite.

Also, a decentralized social network and authenticated transport (AT) protocol is the technology upon which Bluesky is built. Bluesky says its AT Protocol allows users to own their online identity and move their account from one provider to another. According to Bluesky’s blog, users have complete control over their algorithm to give them autonomy over their experience on the app.

The platform temporarily halted new signups over the weekend following Twitter’s ruckus. “We will temporarily pause Bluesky signups while our team continues to resolve the existing performance issues. We’ll keep you updated when invite codes resume functionality. We’re excited to welcome more users to our beta soon,” Bluesky said in a post.

Incidentally, Bluesky is owned by former Twitter CEO Jack Dorsey. Though Bluesky’s architecture is similar to Mastodon’s, many users have found Bluesky to be more intuitive, while Mastodon can come off as inaccessible. However, to remain competitive, Mastodon recently simplified its signup flow, making Mastodon.social the default server for new users.

Hive Social

Like Mastodon and Bluesky, Hive Social was a beneficiary following Musk’s takeover of Twitter late last year. Within one week of Twitter’s change of ownership, Business Insider reports Hive had seen its users spike to more than two million.

According to TechCrunch, Hive Social, launched in 2019, has been dubbed as a Gen Z-focused social app created by Kassandra Pop. Hive combines concepts from various social networks, including Instagram, Twitter, and even MySpace, such as letting users add music to their profiles from Spotify or Apple Music accounts.

Some have argued that out of all alternatives to Twitter, Hive is one of the best. “It is user-friendly, not politically obsessed, and has a neat way of differentiating posts and profile feeds without all the bloat and fluff that modern Twitter has come to possess,” a report reads.

The post Some Twitter alternatives that dodge Musk’s paywall appeared first on TechHQ.

There’s a wave of interest in new ways to discover content online and that includes the use of Web3 architecture such as digital ledgers and decentralization to reshape the search landscape. Entrants – each with their own variation on the theme – include Presearch, Timpi, Kinic, and Kin, to give just a few Web3-enabled examples. But the big question is can blockchain-based search engines beat Google?

READ NEXT

Blockchain security tools safeguard Web3 success

Google has dominated the search engine landscape for years and has proven impossible to displace. Most recently, Neeva shut down its answer engine having failed to convince sufficient numbers of subscribers to sign up. The big challenge – according to the statement given by Neeva co-founders Sridhar Ramaswamy (ex-SVP of Ads at Google) and Vivek Raghunathan (ex-VP of Monetisation at YouTube) – wasn’t building a capable search engine, it was getting users to switch services.

Muscle memory is strong and big-name content discovery tools are deeply embedded in the ecosystems used by millions of internet users. But that’s not to say blockchain-based search engines don’t offer unique advantages. Web3 – the so-called next generation of the internet – has the potential to shift the balance of power away from tech giants such as Google, Microsoft, and other vendors.

Benefits of blockchain-based search engines

Today, changes in search algorithms can make or break websites, and site owners are left guessing as to what’s happening behind the scenes. SEO tools can provide some insight, but far better if search ranking schemes were out in the open and decentralized, putting power back in the hands of users who could upvote pages that served their needs and flag content to be reviewed.

And this highlights some of the transparency that Web3 brings (together with security and privacy, when needed). “If it’s decentralized, anyone can check the source code,” Wyatt Benno, co-founder of Kinic – a search engine built on the Internet Computer Protocol (ICP) – told TechHQ.

Kinic won the Blue Sky category of the Supernova Internet Computer Global Hackathon, which put the blockchain-based search engine project in the sights of VC heavyweight A16Z as well as other partners associated with the event. And some of the clever features of the Web3 search tool include zero-knowledge machine learning (ZKML).

There are clear steps that need to be done to make ZKML more practical. An optimized ZKVM for the ZKML use-case will definitely allow for faster iterations and experimentations.

Have no idea what ZKML is about?
Check out @drCathieSo_eth 's slides https://t.co/ozcma7wVKw https://t.co/BBKwRBVkuc

— Wyatt Benno (@wyatt_benno) June 29, 2023

For example, zero-knowledge technologies give users the ability to prove that they know something without revealing that information, which is useful for building trust without putting sensitive data at risk. The idea for Kinic came about after Benno and his collaborator Houman Shadab had created ICME – a no-code Web3 tool for creating websites and apps that run on blockchain (specifically, the ICP).

Having launched ICME, Benno and Shadab received feedback from users that while it was now much more straightforward to build Web3 projects on ICP, this content – for example, if authors had built a blog – wasn’t easily discoverable. Hence the inspiration for Kinic – their blockchain-based search engine.

READ NEXT

ChatGPT versus Google – the future of search

To picture a website built on ICP, imagine a smart contract (or ‘Frontend Container’ in ICP speak) that contains the site’s HTML and CSS code. The URL could be the canister ID and potentially flag whether any changes had been made to the site.

Benno has a long list of ways that Web3 architecture can benefit app owners and users. Digital ledgers can validate that content belongs to a particular website – for example, to help readers determine whether news articles are genuine or have been cut and pasted from elsewhere. And the decentralized architecture makes legitimate information harder for authorities to suppress.

Picking up on the theme of decentralized computing – noting a recent story on TechHQ – he points out that it would be possible for GPU hosts to prove to customers that data has been processed as instructed. And this leads the discussion onto arguably one of the biggest ways that blockchain-based search engines could beat Google.

Selling your search skills

“Users could make their own machine learning search models and sell them,” Benno comments. “Owners could set the price and keep the weights hidden.” Today, major search engine providers benefit greatly from user data. But in the future, those treasure troves of search queries – for example, if used to fine-tune machine learning models – could generate paydays for users rather than big tech.

Benno has had interest from VC firms wanting to know which topics are trending on Web3 search to give them a head start in discovering early-stage, high-growth opportunities. And API access to URL click counters could be one way to monetize Kinic, while staying true to its principles of not saving or sharing any personal information.

SnapFusion: Text-to-Image Diffusion Model on Mobile Devices within Two Seconds

paper page: https://t.co/ykIrW1u1Uo

present a generic approach that, for the first time, unlocks running text-to-image diffusion models on mobile devices in less than 2 seconds. We achieve so by… pic.twitter.com/rJjPjCjRDw

— AK (@_akhaliq) June 2, 2023

Big companies are looking at Web3 opportunities with interest – for example, Starbucks is testing the idea of a digital rewards program using NFTs. And with smartphones capable of running efficiently fine-tuned generative AI models, as well as the emergence of cheap-to-produce LLM updates, it’s an interesting time in search land and its related technologies.

The post Can blockchain-based search engines beat Google? appeared first on TechHQ.

• Z-Library alternatives growing, opening users up to scams.
• Z-Library deploying its own new access method.
• Z-Library embroiled in a complex legal case.

Z-Library, the shadow library that allows file sharing of academic journals, is finding alternative ways to stay alive despite significant attacks and losses in recent months.

A new access method has been launched by the shadow library to help improve accessibility. The dedicated desktop application will make it easier to access the site going forward, after several rounds of domain seizures by the US government.

READ NEXT

Why it’s time to give staff open access to first-party data

One of the reasons for the development of this software is the criminal case that Z-Library is embroiled in. In November 2022, Z-Library lost access to over 200 domain names after two of its alleged operators were arrested in Argentina. Both defendants, who are Russian, retained US lawyers to fight their cases.

However, both Valeriia Ermakova – who has hired the services of Temkin & Associates – and Anton Napolsky – who is being represented by Brown Legal Consulting – have opted for lawyers who are fluent in Russian.

More recently, there’ve been more domain seizures. Last year, Z-Library approached the issue by initially denying that it had been targeted, but this time around the site’s operators were quick to confirm the action, directing users to alternative login screens through a Telegram message.

“Unfortunately, one of our primary login domains was seized today. Therefore, we recommend using the domain singlelogin.re to log in to your account, as well as to register,” the Z-Library team wrote.

Via TorrentFreak.com

Z-Library alternatives gain traffic

Although the shadow library has pushed ahead through recriminations, lost domain names send traffic to knockoffs and alternatives. Z-Library alternatives are now getting millions of visitors a month, but don’t have the same track-record as the original, putting users at risk of scams.

The shadow library also faced a huge bot attack earlier this month, causing technical issues: registrations stopped working and email delivery was interrupted. These are also likely factors in the development of the new desktop software.

The desktop launcher will be available for Mac, Windows and Linux platforms, and will automatically redirect users to the right place without relying on a single domain name. Previously, users accessed Z-Library through a dedicated URL that directed them to a ‘personal’ domain, which provided access to Z-Library.

The use of subdomains was working, but could easily have been wiped out by yet another round of domain seizures. The team announced that the launcher “will save you the trouble of searching for a working website link, as it will handle everything for you.”

Not only does it simplify access to Z-Library, but it can connect over the Tor network, which helps evade blocking efforts and adds another layer of privacy. Apparently, the software will likely trigger a notice that it’s from an unverified developer; Z-Library says this is standard, but of course users should treat third-party applications with caution.

The accessibility argument

It might be unexpected that the service would plough ahead during an active criminal case. In this respect, its reminiscent of how The Pirate Bay positioned itself years ago. The Z-Library team sees “free access to literature” as a main driver.

YOU MIGHT LIKE

COMPLIANCE

A win for open access as Z-library goes physical

“The goal of Z-Library is to provide free access to literature to as many people in need as possible. Books are the scientific and cultural heritage of all humankind, and we strive to preserve this legacy and use its power for the benefit of our society.”

“We don’t promote piracy. The work of authors and publishers should be paid for and valued,” the Z-Library team explains, adding that it supports copyright legislation and doesn’t aim to change any laws.

However, free access to literature is paramount to many students’ studies, particularly in remote or underfunded areas.

Source: https://twitter.com/ApalaBhowmick/status/1572613014444212225?s=20

According to Bhowmick (the above tweet’s author), the Z-library shutdown is another pattern of racism and inequity that hinders promising young people from following their academic passions to futures in wealthier countries with more supported academic institutions. “There are scant ways of finding access to scholarly literature in India due to fractured print culture networks and limited incomes,” she said, “It’s almost a deliberate strategy to gate-keep academia from those who are racialized, or marginalized in other ways, especially in vulnerable economies in the world.”

Never underestimate the power of free access to literature.

The post Z-Library finds alternative ways to stay active appeared first on TechHQ.