Corsight plans to release a new product that combines DNA and face recognition technology and could have significant law enforcement and privacy implications.
In this report, we examine Corsight’s product roadmap for “DNA to FACE,” presented at the 2021 Imperial Capital Investors Conference, possible use cases for the technology, and warnings from a privacy expert.
IPVM collaborated with MIT Technology Review on this report, see the MIT Technology Review article: This company says it’s developing a system that can recognize your face from just your DNA Read More
Monthly Archives: January 2022
The new version of GPT-3 is much better behaved (and should be less toxic)
OpenAI has built a new version of GPT-3, its game-changing language model, that it says does away with some of the most toxic issues that plagued its predecessor. The San Francisco-based lab says the updated model, called InstructGPT, is better at following the instructions of people using it—known as “alignment” in AI jargon—and thus produces less offensive language, less misinformation, and fewer mistakes overall—unless explicitly told not to do so.
… Previous attempts to tackle the problem included filtering out offensive language from the training set. But that can make models perform less well, especially in cases where the training data is already sparse, such as text from minority groups.
The OpenAI researchers have avoided this problem by starting with a fully trained GPT-3 model. They then added another round of training, using reinforcement learning to teach the model what it should say and when, based on the preferences of human users. Read More
Introducing the First Self-Supervised Algorithm for Speech, Vision and Text
- We’re introducing data2vec, the first high-performance self-supervised algorithm that learns in the same way for speech, vision and text.
- With data2vec, we’re closer to building machines that learn about different aspects of the world around them without having to rely on labeled data. Read More
Three things web3 should fix in 2022
A viral video highlights some very real shortcomings in the next-generation internet
Last weekend, it felt like everyone I knew was sending me the same link. “The Problem With NFTs,” a long video essay by the Canadian media critic Dan Olson, ricocheted around all corners of the tech world since it was uploaded on Friday. (It now has 2.6 million views and climbing.) Over 138 meticulously researched minutes, Olson traces the history of the 2008 financial crisis, the creation of Bitcoin and Ethereum, and the rise of NFTs and DAOs, and reaches the conclusion that what we have taken to calling “web3” is effectively beyond saving: the technology is too broken, and its creators too indifferent to its failures, for it to ever to live up to the promise of its most starry-eyed backers.
Few of Olson’s criticisms are entirely new, and on my Twitter timeline this week, I saw many crypto enthusiasts dismiss them out of hand. Few people working in the space will be surprised to learn that crypto3 is awash in grifts; that current blockchains are energy inefficient and expensive; or that digital wallets are difficult to use and fraught with danger. Many web3 builders will also bristle at Olson’s tone, which is smug and hectoring in the house style of the YouTube video essayist; his audience is not people working in crypto, but rather everyone he thinks ought to be afraid of those people.
And yet the collective force of Olson’s arguments is substantial. His essay explains the rise of cryptocurrencies through the lens of rising inequality; pandemic-era isolation and loneliness; self-dealing venture capitalists; and a desperate sense among young strivers that the future is only ever getting smaller. All of which feels particularly timely, given this week’s crash in crypto prices. Read More
The False Promise of Web3
I didn’t anticipate how much I’d appreciate a @Jack of fewer trades. Key to progress is class traitors: Generals warning of a military-industrial complex, product managers who narc on mendacious management, and tech leaders who violate the Silicon Valley code of the white guy — never criticize each other or your noble missions to save the world. Jack Dorsey has drawn his sword and taken aim at colleagues attempting to centralize control and gain from the promise of decentralization. Specifically, “web3.”
What is web3? It’s a hazy/vague term describing a crypto-powered internet, and a font of yogababble. Its promoters would say something akin to:
Web3 is a decentralized version of the internet where platforms and apps are built and owned by users. Unlike web2 (the current web), which is dominated by centralized platforms such as Google, Apple, and Facebook, web3 will use blockchain, crypto, and NFTs to transfer power back to the internet community.
Sounds good. Most of us buy the down-with-Facebook-and-Google narrative. Cut out the middleman, and we all win — especially the little guys. The dispersion of theaters, doctors offices, and bank branches to our homes, smart speakers, and phones offers enormous potential to provide elemental services with reduced friction. Smart contracts could, among other things, reduce agency costs and iron systemic biases out of the process. That’s the promise of decentralization. But does the music match the words? Read More
The new version of GPT-3 is much better behaved (and should be less toxic)
OpenAI has built a new version of GPT-3, its game-changing language model, that it says does away with some of the most toxic issues that plagued its predecessor. The San Francisco-based lab says the updated model, called InstructGPT, is better at following the instructions of people using it—known as “alignment” in AI jargon—and thus produces less offensive language, less misinformation, and fewer mistakes overall—unless explicitly told not to do so.
Large language models like GPT-3 are trained using vast bodies of text, much it taken from the internet, in which they encounter the best and worst of what people put down in words. That is a problem for today’s chatbots and text-generation tools. The models soak up toxic language—from text that is racist and misogynistic or that contains more insidious, baked-in prejudices—as well as falsehoods.
OpenAI has made IntructGPT the default model for users of its application programming interface (API)—a service that gives access to the company’s language models for a fee. GPT-3 will still be available but OpenAI does not recommend using it. “It’s the first time these alignment techniques are being applied to a real product,” says Jan Leike, who co-leads OpenAI’s alignment team.
Previous attempts to tackle the problem included filtering out offensive language from the training set. But that can make models perform less well, especially in cases where the training data is already sparse, such as text from minority groups.
The OpenAI researchers have avoided this problem by starting with a fully trained GPT-3 model. They then added another round of training, using reinforcement learning to teach the model what it should say and when, based on the preferences of human users. Read More
Introducing: Macaw
Macaw is a high-performance question-answering (QA) model capable of outperforming other popular current language models, all while being an order of magnitude smaller. This demo allows you to explore Macaw’s answers and compare them to those of the popular GPT-3 language model on a benchmark set of questions. Read More
Code
TransformerFusion
Monocular RGB Scene Reconstruction using Transformers
We introduce TransformerFusion, a transformer-based 3D scene reconstruction approach. From an input monocular RGB video, the video frames are processed by a transformer network that fuses the observations into a volumetric feature grid representing the scene; this feature grid is then decoded into an implicit 3D scene representation. Key to our approach is the transformer architecture that enables the network to learn to attend to the most relevant image frames for each 3D location in the scene, supervised only by the scene reconstruction task. Features are fused in a coarse-to-fine fashion, storing fine-level features only where needed, requiring lower memory storage and enabling fusion at interactive rates. The feature grid is then decoded to a higher-resolution scene reconstruction, using an MLP-based surface occupancy prediction from interpolated coarse-to-fine 3D features. Our approach results in an accurate surface reconstruction, outperforming state-of-the-art multi-view stereo depth estimation methods, fully-convolutional 3D reconstruction approaches, and approaches using LSTM- or GRU-based recurrent networks for video sequence fusion. Read More
Mobile-Former: Bridging MobileNet and Transformer
We present Mobile-Former, a parallel design of MobileNet and transformer with a two-way bridge in between. This structure leverages the advantages of MobileNet at local processing and transformer at global interaction. And the bridge enables bidirectional fusion of local and global features. Different from recent works on vision transformer, the transformer in Mobile-Former contains very few tokens (e.g. 6 or fewer tokens) that are randomly initialized to learn global priors, resulting in low computational cost. Combining with the proposed light-weight cross attention to model the bridge, Mobile-Former is not only computationally efficient, but also has more representation power. It outperforms MobileNetV3 at low FLOP regime from 25M to 500M FLOPs on ImageNet classification. For instance, Mobile-Former achieves 77.9% top-1 accuracy at 294M FLOPs, gaining 1.3% over MobileNetV3 but saving 17% of computations. When transferring to object detection, Mobile-Former outperforms MobileNetV3 by 8.6 AP in RetinaNet framework. Furthermore, we build an efficient end-to-end detector by replacing backbone, encoder and decoder in DETR with Mobile-Former, which outperforms DETR by 1.1 AP but saves 52% of computational cost and 36% of parameters. Read More
#image-recognitionMeta has built an AI supercomputer it says will be world’s fastest by end of 2022
Social media conglomerate Meta is the latest tech company to build an “AI supercomputer” — a high-speed computer designed specifically to train machine learning systems. The company says its new AI Research SuperCluster, or RSC, is already among the fastest machines of its type and, when complete in mid-2022, will be the world’s fastest.
… The news demonstrates the absolute centrality of AI research to companies like Meta. Rivals like Microsoft and Nvidia have already announced their own “AI supercomputers,” which are slightly different from what we think of as regular supercomputers. RSC will be used to train a range of systems across Meta’s businesses: from content moderation algorithms used to detect hate speech on Facebook and Instagram to augmented reality features that will one day be available in the company’s future AR hardware. And, yes, Meta says RSC will be used to design experiences for the metaverse — the company’s insistent branding for an interconnected series of virtual spaces, from offices to online arenas. Read More
Rick's Cafe AI 