Technology
This week in AI: Generative AI spams academic journals
Hello guys, welcome to TechCrunch’s regular AI newsletter.
This week in AI, generative AI is beginning to spam academic publications, a discouraging recent development on the disinformation front.
IN post about retraction watchblog tracking the recent retreat from academic research, assistant professors of philosophy Tomasz Żuradzk and Leszek Wroński wrote about three journals published by Addleton Academic Publishers that appear to consist entirely of AI-generated articles.
Magazines feature articles following the identical template, stuffed with buzzwords like “blockchain,” “metaverse,” “internet of things,” and “deep learning.” They list the identical editorial committee – 10 members of that are deceased – and an not easily seen address in Queens, New York, that appears to be home.
So what is going on on? You can ask. Isn’t viewing AI-generated spam simply a value of doing business online lately?
Yeah. But the fake journals show how easily the systems used to guage researchers for promotions and tenure might be fooled – and this could possibly be a motivator for knowledge employees in other industries.
In a minimum of one widely used rating system, CiteScore, journals rank in the highest ten for philosophy research. How is it possible? They quote one another at length. (CiteScore includes citations in its calculations). Żuradzk and Wroński state that of the 541 citations in considered one of Addleton’s journals, 208 come from other false publications of the publisher.
“(These rankings) often serve as indicators of research quality for universities and funding institutions,” Żuradzk and Wroński wrote. “They play a key role in decisions about academic rewards, hiring and promotion, and thus can influence researchers’ publication strategies.”
You could argue that CiteScore is the issue – it’s clearly a flawed metric. And this just isn’t a false argument. However, it just isn’t incorrect to say that generative AI and its abuse are disrupting the systems on which individuals’s lives depend in unexpected – and potentially quite harmful – ways.
There is a future in which generative AI forces us to rethink and redesign systems like CiteScore to be more equitable, holistic and inclusive. The bleaker alternative – and the one which exists today – is a future in which generative AI continues to run amok, wreaking havoc and ruining work lives.
I hope we’ll correct course soon.
News
DeepMind soundtrack generator: DeepMind, Google’s artificial intelligence research lab, says it’s developing artificial intelligence technology to generate movie soundtracks. DeepMind’s AI combines audio descriptions (e.g. “jellyfish pulsating underwater, sea life, ocean”) with the video to create music, sound effects, and even dialogue that match the characters and tone of the video.
Robot chauffeur: : Researchers on the University of Tokyo have developed and trained a “musculoskeletal humanoid” named Musashi to drive a small electric automobile around a test track. Equipped with two cameras replacing human eyes, Musashi can “see” the road ahead and the views reflected in the automobile’s side mirrors.
New AI search engine: : Genspark, a brand new AI-powered search platform, uses generative AI to create custom summaries in response to queries. To date, $60 million has been raised from investors including Lanchi Ventures; In its latest round of financing, the corporate valued it at $260 million post-acquisition, which is decent considering Genspark competes with rivals like Perplexity.
How much does ChatGPT cost?: How much does ChatGPT, OpenAI’s ever-expanding AI-powered chatbot platform, cost? Answering this query is tougher than you think that. To keep track of different ChatGPT subscription options available, we have prepared an updated ChatGPT pricing guide.
Science article of the week
Autonomous vehicles face an infinite number of edge cases, depending on location and situation. If you are driving on a two-lane road and someone activates their left turn signal, does that mean they’ll change lanes? Or that it is best to pass them on? The answer may depend upon whether you are driving on I-5 or the highway.
A gaggle of researchers from Nvidia, USC, UW and Stanford show in a paper just published in CVPR that many ambiguous or unusual circumstances might be solved by, should you can consider it, asking an AI to read the local drivers’ handbook.
Their Big Tongue Driver Assistant or LLaDa, gives the LLM access – even without the power to specify – driving instructions for a state, country or region. Local rules, customs or signs might be found in the literature, and when an unexpected circumstance occurs, e.g. a horn, traffic lights or a flock of sheep, an appropriate motion is generated (stop, stop, honk).
This is certainly not an entire, comprehensive driving system, nevertheless it shows another path to a “universal” driving system that also encounters surprises. Or perhaps it is a way for the remainder of us to search out out why people honk at us after we visit unknown sites.
Model of the week
On Monday, Runway, an organization that creates generative AI tools aimed toward creators of film and image content, presented Gen-3 Alpha. Trained on an enormous variety of images and videos from each public and internal sources, Gen-3 can generate video clips based on text descriptions and still images.
Runway claims Gen-3 Alpha delivers “significant” improvements in generation speed and fidelity over Runway’s previous video flagship, Gen-2, in addition to precise control over the structure, style and motion of the videos it creates. Runway says Gen-3 can be customized to offer more “stylistically controlled” and consistent characters, guided by “specific artistic and narrative requirements.”
The Gen-3 Alpha has its limitations – including the indisputable fact that footage lasts a maximum of 10 seconds. But Runway co-founder Anastasis Germanidis guarantees that that is just the primary of several video-generating models to come back in a family of next-generation models trained on Runway’s improved infrastructure.
Gen-3 Alpha is the most recent of several generative video systems to hit the scene in recent months. Others include OpenAI’s Sora, Luma’s Dream Machine, and Google’s Veo. Together, they threaten to upend the film and tv industry as we understand it – assuming they will overcome copyright challenges.
Take the bag
AI won’t take your next McDonald’s order.
McDonald’s this week announced that it should remove automated order-taking technology, which the fast food chain has been testing for the higher a part of three years, from greater than 100 of its restaurants. The technology — developed with IBM and installed in drive-thru restaurants — became popular last 12 months due to its tendency to misunderstand customers and make mistakes.
Recent piece in Takeout suggests that artificial intelligence is losing its grip on fast food operators, who’ve recently expressed enthusiasm for the technology and its potential to extend efficiency (and reduce labor costs). Presto, a significant player in the AI-powered drive-thru lane market, recently lost a significant customer, Del Taco, and is facing mounting losses.
The problem is inaccuracy.
McDonald’s CEO Chris Kempczinski he said CNBC in June 2021 found that its voice recognition technology was accurate about 85% of the time, but that it needed to be assisted by human staff for about one in five orders. Meanwhile, in keeping with Takeout, one of the best version of the Presto system processes only about 30% of orders without human assistance.
So, so long as artificial intelligence is there decimating some segments of the gig economy appear to think that certain jobs – especially people who require understanding a wide range of accents and dialects – can’t be automated. At least for now.