Last Week in AI #129: AI to write code, AI for nuclear fusion, and more!

OpenAI's new Codex can translate language prompts to code, AI is being applied to many problems in nuclear fusion R&D, and other news stores from last week.

Aug 16, 2021

OpenAI can translate English into code with its new machine learning software Codex

Last week OpenAI demoed an impressive application of its AI language research. The new OpenAI Codex is able to translate natural language commands into valid and functional code. The live YouTube demo showed Codex being used to build a simple Javascript-based game and code small snippets to interact with programs with APIs, like Microsoft Word. OpenAI describes Codex as something that can help programmers translate small logical procedures into code, but the programmer is still needed to break down a larger problem into these smaller chunks. Successful future developments along this line will help improve programmers’ productivity and make it easier to code. As OpenAI puts it:

Codex is bringing computers closer to humans, letting them speak English rather than machine code

Can AI Make a Better Fusion Reactor?

Nuclear fusion holds many promises for the future of clean and sustainable energy. Its development however, has been marred by delays and technical difficulties. Researchers are now trying to adopt tools from machine learning to help the process. For example, AI models can be trained to detect and forecast disruptions in plasma that needs to be stabilized for sustained fusion reactions. Other researchers are focusing on applying AI to speed up complex calculations. One model was trained to generate complex reaction visualizations in seconds, as opposed to the minutes required by traditional methods. Another was used to compute the electric field generated by plasma for research. With these developments, experts note that "AI and machine learning will be a key tool" toward bringing fusion to reality.

Twitter AI bias contest shows beauty filters hoodwink the algorithm

The results of Twitter's first algorithmic bias bounty contest were revealed last week. The contest challenged participants to find issues with Twitter's AI-based auto-cropping algorithm, which was criticized by users last year for seemingly favoring white faces over other races. The contest, the first of its type, was an experiment in applying the concept of 'bug bounties' -- payments issued by companies to external teams that find flaws in software -- to the context of AI algorithms. The winning contestant, researcher Bogdan Kulynych at Switzerland's EPFL technical university won a $3,500 prize for determining that the algorithm favors "faces that look slim and young and with skin that is lighter-colored or with warmer tones."

If you are not subscribed yet and like our AI news digests, feel free to subscribe!

Other News

Research

Researchers Create 'Master Faces' to Bypass Facial Recognition - "In their paper, researchers at the Blavatnik School of Computer Science and the School of Electrical Engineering in Tel Aviv detail how they successfully created nine "master key" faces that are able to impersonate almost half the faces in a dataset of three leading face recognition systems."

New artificial intelligence tool helps forecast Amazon deforestation - "Nearly 10,000 square kilometers of the Brazilian Amazon, an area the size of Lebanon, is at high risk of being cleared, according to a new tool using artificial intelligence technology to help forecast deforestation before it actually happens."

Without Code for DeepMind’s Protein AI, This Lab Wrote Its Own - "For biologists who study the structure of proteins, the recent history of their field is divided into two epochs: before CASP14, the 14th biennial round of the Critical Assessment of Protein Structure conference, and after."

This New Study Shows Artificial Neural Networks (ANN) Based On Human Brain Connectivity Can Perform Cognitive Tasks Efficiently - "Possibly a new breakthrough has been achieved in the domain of artificial intelligence."

Google AI Introduces 'SoundStream', A Neural Audio Codec That Provides High-Quality Audio For Various Sound Classes, Improving Machine Learning-Driven Audio Codecs - "Audio codecs are tools that compress your sound files to make them smaller and less time-consuming. They are essential for streaming, as they save you from having to use up so much data on the internet while you listen in."

Artificial Intelligence may diagnose dementia in a day - "Scientists are testing an artificial-intelligence system thought to be capable of diagnosing dementia after a single brain scan. It may also be able to predict whether the condition will remain stable for many years, slowly deteriorate or need immediate treatment."

Applications

Pizza Hut is building an AI that it says will recommend food to you based on your local weather - "Pizza Hut plans to use AI tech to recommend orders to customers based on the weather in their location, the company's data chief told VentureBeat."

New Google project in Israel aims to use AI to help detect colon cancer - "Google’s Verily Life Sciences has said it will open a research and development center in Israel that will focus on applying artificial intelligence technology to biomedical problems."

Cattle-ist for the Future: Plainsight Revolutionizes Livestock Management with AI - "Computer vision and edge AI are looking beyond the pasture. Plainsight, a San Francisco-based startup and NVIDIA Metropolis partner, is helping the meat processing industry improve its operations — from farms to forks."

Business

This hot startup is now valued at $1 billion for its A.I. skills - "Snorkel AI, a startup focussed on data labeling, landed $85 million in funding and is now valued at $1 billion."

TikTok brain drain as head of AI lab leaves ByteDance? - "In May, ByteDance founder and CEO Zhāng Yīmíng 张一鸣 announced that he would step down from his chief executive position."

China's tech crackdown thwarts Pony.ai's U.S. listing plans -sources - " Autonomous driving startup Pony.ai has put on hold plans to go public in New York through a merger with a blank-check firm at a $12 billion valuation, after it failed to gain assurances from Beijing that it would not become a target of a crackdown against Chinese technology companies, people familiar with the matter said."

Why Microsoft and Twitter are using bug bounties to fix A.I. - "Companies like Microsoft and Twitter have been increasingly hosting bug bounty programs to entice well-meaning hackers to spot flaws in their software."

Doxel raises $40M from Insight, a16z to become the ‘Waze for construction’ - "Doxel, which has developed software that uses computer vision to help track and monitor progress on construction job sites, announced today that it has raised $40 million in Series B funding."

AI21 Labs trains a massive language model to rival OpenAI’s GPT-3 - "For the better part of a year, OpenAI’s GPT-3 has remained among the largest AI language models ever created, if not the largest of its kind."

Concerns

5 Teslas Have Crashed On The Same Road In California - "Sharing his story on Reddit and going by the name BBFLG, he sent his post with the intention of Tesla improving the technology. Apparently, BBFLG says that five Teslas including his Model X crashed in the same spot on Northside Drive in Yosemite Valley, California."

U.S. prisons mull AI to analyze inmate phone calls - "For people like Heather Bollin, a 43-year-old woman in Texas engaged to a man who is currently incarcerated, constant surveillance is a fact of life: the three daily phone calls they have together are subject to monitoring by prison officials."

Deleting unethical data sets isn’t good enough - "The AI research community has tried to scrub away its past. But the internet is forever In 2016, hoping to spur advancements in facial recognition, Microsoft released the largest face database in the world. Called MS-Celeb-1M, it contained 10 million images of 100,000 celebrities’ faces."

AI can persuade people to make ethically questionable decisions, study finds - "In a large-scale survey leveraging OpenAI’s GPT-2 language model, the researchers found AI’s advice can “corrupt” people even when they’re aware the source of the advice is AI."

Google is developing a new superintelligent AI but ethical questions remain - "Google is developing a nimble, multi-purpose AI that can perform millions of tasks. Called Pathways, Google’s solution seeks to centralize disparate AI into one powerful, all-knowing algorithm."

Analysis

Appreciating the Poetic Misunderstandings of A.I. Art - "What does an “Art Deco Buddhist temple” look like? The phrase is nearly nonsensical; it’s hard to imagine a Buddhist temple built in the Art Deco style, the early-twentieth-century Western aesthetic of attenuated architecture and streamlined forms."

Expert Opinions

Kai-Fu Lee : automation may mean workplaces preserve 'the pretense of working' - "With fully-automated entry-level work, companies lose an opportunity to recognize and groom talent."

Podcast

Check out our weekly podcast covering these stories!
Website | RSS | iTunes | Spotify | YouTube

Last Week in AI