Stanford’s Alpaca AI performs similarly to the astonishing ChatGPT on many tasks – but it’s built on an open-source language model and cost less than US$600 to train up. It seems these godlike AIs are already frighteningly cheap and easy to replicate.
It’s no secret that OpenAI’s ChatGPT has some incredible capabilities—for instance, the chatbot can write poetry that resembles Shakespearean sonnets or debug code for a computer program. These abilities are made possible by the massive machine-learning model that ChatGPT is built upon. Researchers have found that when these types of models become large enough, extraordinary capabilities emerge.
But bigger models also require more time and money to train. The training process involves showing hundreds of billions of examples to a model. Gathering so much data is an involved process in itself. Then come the monetary and environmental costs of running many powerful computers for days or weeks to train a model that may have billions of parameters.
“It’s been estimated that training models at the scale of what ChatGPT is hypothesized to run on could take millions of dollars, just for a single training run. Can we improve the efficiency of these training methods, so we can still get good models in less time and for less money? We propose to do this by leveraging smaller language models that have previously been trained,” says Yoon Kim, an assistant professor in MIT’s Department of Electrical Engineering and Computer Science and a member of the Computer Science and Artificial Intelligence Laboratory (CSAIL).
Presented by Intel.
Every day around the world, companies leverage artificial intelligence to accelerate scientific discovery, and transform consumer and business services. Regrettably, the employment of AI is not occurring evenly. McKinsey’s ‘The State of AI in 2022’ report documents that adoption of AI by organizations has stalled at 50%. AI leaders are pulling ahead of the pack. One reason is 53% of AI projects fail to get to production. As the benefits of AI to everyone are too great and the issues with AI being in the hands of only a few are too concerning, that it is an opportune time to survey the challenges of going from concept to deployment.
Chipmaker Nvidia has announced a new service that will allow users to access supercomputer level computing power from web browsers. The service, called Nvidia DGX Cloud, is an AI supercomputing service that allows enterprises to run workloads on the company’s A100 and H100 chips remotely.
“DGX Cloud provides dedicated clusters of NVIDIA DGX AI supercomputing, paired with NVIDIA AI software. The service makes it possible for every enterprise to access its own AI supercomputer using a simple web browser, removing the complexity of acquiring, deploying and managing on-premises infrastructure,” the company said in a blog post.
As reported by Reuters, however, the service isn’t cheap. Nvidia is charging $37,000 per month for access to eight of the A100 or H100 chips — the company’s flagship chips, which are both designed for AI computing. “Each instance of DGX Cloud features eight Nvidia H100 or A100 80GB Tensor Core GPUs for a total of 640GB of GPU memory per node. A high-performance, low-latency fabric built with Nvidia Networking ensures workloads can scale across clusters of interconnected systems, allowing multiple instances to act as one massive GPU to meet the performance requirements of advanced AI training,” Nvidia’s blog post explained.
GitHub Copilot is also coming to pull requests to help developers create AI-generated descriptions. Tags are automatically completed by GitHub Copilot based on what code has changed, and developers can then review and edit them.
“At GitHub we invented the pull request over a decade ago, so the natural next step for us was to bring Copilot into the pull request,” says Dohmke. “You can actually ask Copilot to describe the pull request to you, or you can ask Copilot to generate tests.”
Electronic neurons made from silicon mimic brain cells and could be used to treat autism1.
Researchers plan to use the technology in conjunction with machine learning to retrain damaged or atypical neurons and restore function in the brains of people with Alzheimer’s disease, autism or other conditions.
Another team attempted to make artificial neurons in 2015 from a conductive organic chemical, but that version oversimplified brain signaling and was too large to implant in a human brain2.
GitHub is announcing its Copilot X initiative today, an extension of its work on its popular Copilot code completion tool, which originally launched into preview all the way back in 2021. With this, the Microsoft-owned company is launching a code-centric chat mode for Copilot that helps developers write and debug their code, as well as Copilot for pull requests, AI-generated answers about documentation and more. Unsurprisingly, these new features are powered by OpenAI’s GPT-4, though it’s worth noting that, mostly for latency reasons, the code completion tool remains on GitHub’s Codex model, which it derived from GPT-3.
“With the new model coming online, we asked ourselves: what’s the next step? What’s the next step for Copilot? We believe that for auto completion, we nailed that scenario,” GitHub CEO Thomas Dohmke told me.
“How do we mitigate the downside of technology, to make sure that human values, public interest and democracy are built into the system?”
Mozilla, the not-for-profit force behind the Firefox browser, is launching an AI-focused startup with a mission to create an open source and trustworthy alternative to emerging heavyweights like ChatGPT. The company this morning announced that Moez Draief’, a former global chief scientist with Capgemini Invent, will head the venture, which has a $30 million seed investment from Mozilla Foundation.
Mozilla Foundation president Mark Surman spoke with Forbes about the new venture, called Mozilla.
What Mozilla did for browsers with Firefox, it’s now planning to do in the realm of AI. Foundation chief Mark Surman says there’s already momentum to take on the titans of tech.
This video is about the 2022 robot exhibition that was done in tokyo.
More of our videos.
HOW TO INSTALL THE ARDUINO IDE
HOW TO BUILD AN ARDUINO VEHICLE #arduino keke.
https://www.youtube.com/watch?v=brV8IQynVgs.
ROBOT LEG
Scientists in the US managed to put together a living computer by cultivating over 80,000 mouse stem cells (opens in new tab) (via IT Home) (opens in new tab). One day, the hope is to have a robot that uses living muscle tissue to sense and process information about its environment.
Researchers at the University of Illinois have used tens of thousands of living mouse brain cells to build a computer that can recognize patterns of light and electricity. The team presented their findings at the American Institute of Physics in the form of a computer about the size of your palm.
