Tech giants are open-washing their artificial intelligence models
How fast can a robot get? Case study of a bot solving a Rubik’s Cube in a Jammed World
The market capitalization of chipmaker Nvidia amid a surge in demand for computer chips that power AI models — making it the most valuable company in the world. 3 min read by The Guardian
Meta has expanded its translation model to 204 languages, with 150 being low-resource ones. These are tongues such as Tswana, Samoan or Gaelic, for which there’s not enough data to train a machine-translation system. The Meta model was fed with translations mined from the internet and around 6,000 ‘seed’ sentences created by professional translators. Although the translations are “not very good yet”, says Gaelic linguist William Lamb, he applauds the team’s effort. He said that if they want to improve the translation, they need to speak to the people who are still living and breathing the language.
A chatbot can check another’s workings by asking it the same question multiple times and assessing how much the responses differ from one another. The output isn’t reliable because of a high degree of randomness. Sebastian says he might ask you the same question over and over again in order to check if you are just making things up. If you give a different answer every time, something isn’t right. Two human judges who looked at 100 of the first chatbot’s answers agreed with the ‘judge’ chatbot 93% of the time.
You will miss it if you blink. Mitsubishi’s TOKUFASTbot solves a Rubik’s Cube in around a quarter of a second — a new world record for a robot. An artificial intelligence program identifies the cube’s colors, particularly the difficult-to-Distinguish red and orange ones, and then produces the shortest rotation sequence to a solution. The cube became jammed during the first record attempt because the robot’s motors were moving so fast. (Smithsonian Magazine | 4 min read)
Artificial Intelligence as a New Tool for Self-Driving Vehicles: An Editorial on Openness and the Safety of Autonomous Vehicles
According to an Nature Biomedical Engineering editorial, specialist artificial intelligence systems may soon offer most of the feedback given during traditional peer review. Human reviewers would no longer have to worry about high level conceptual and technical feedback. AI agents might even become the go-to interface for reading papers with each paper having its own persistent ‘author chatbot’ ready to answer any queries, the editorial suggests. Problems such as fake data and plagiarism should be solved first by the community.
Proponents of open source say the models used to power the systems are not open source but are instead derived from commercial software that is freely available. Advocates say that full open source is crucial in the effort to make an accountable artificial intelligence model. Despite the claims of openness, a lot of companies still don’t give out important information.
Statistics suggest that autonomous vehicles have fewer accidents than human drivers — but limited data is hindering attempts to understand the safety of self-driving cars, say researchers. They analysed thousands of accidents and found almost 550 in which people and autonomous vehicles crashed in similar circumstances. During turns, the technology crashed twice as often as humans and, at dusk or dawn, five times more often. There is no comprehensive assessment of the safety of self-driving car technology due to the low number of crashes available to analyse.
Why AIs could be anything but, the derailment risks of long freight trains, and breeding better wheat: A study with a genetic engineering team
Some people on the spectrum are turning to online companions such as Replika, Paradot, and others in order to find better connections. “These interactions give me more confidence when talking to real humans because they’ve helped me to try certain conversation skills that can be applied in real life,” says Elías López, who was diagnosed as autistic at the age of 30. So far, there has been little research into whether these chatbots have therapeutic benefits. Some mental-healthcare professionals are concerned that the practice could exacerbate users’ isolation. Catherine Lord, a clinical psychologist, says that it’s helpful to know what risks are and what true value is.
There are no federal restrictions on the length of a freight train in the US and as companies look for longer locomotives, questions arise about whether they are at greater risk of derailing. To find out, a team analysed data on accidents to predict the chances of longer trains coming off the tracks. Replacing two 50-car freight trains with a single 100 car train raises the odds of a train going off the tracks by 11 percent and increases the chance of a train getting stuck. Economic pressures may lead the freight industry to experiment with longer trains.
Genes from century-old wheat varieties could be used to breed useful traits into modern crops, helping them become more disease tolerant and reducing their need for fertilizer. Researchers sequenced the genomes of hundreds of historic varieties of wheat held in a seed collection from the 1920s and ’30s, revealing a huge amount of genetic diversity unseen in modern crops. Plant breeding enabled the team to identify some of the areas of the plants’ genomes responsible for traits such as nutritional content and stress tolerance. In the long run, this knowledge could be used to improve modern wheat varieties.
NaturePodcast: An Efficient Environment for Non-Standard Audio and Streaming Services in Deep Inelastic Scattering
Do not miss an episode. You can listen to the NaturePodcast on Apple Podcasts, as well as on other audio and streaming services. The NaturePodcast has an RSS feed as well.
