Welcome to our roundup of this week’s AI news. If you’re reading this then the robots haven’t taken over yet.
This week AI aced a maths test.
AI fakes made us think everyone is lying.
And artists are using poison to fight data scrapers.
Let’s dig in.
Google takes shape
Our 2024 wishlist got off to a great start this week when Google gave us a peek at Lumiere, its text-to-video model. The video generation and editing in the demo look incredible.
Less flashy, but perhaps more impressive, was Google DeepMind’s unveiling of its AI math whiz called AlphaGeometry. International Mathematical Olympiad geometry problems are notoriously tough to solve, but AlphaGeometry nearly aced them all.
Even AlphaGeometry may find it a challenge to count how many tabs I’ve got open on my browser. Google gave Chrome a revamp with three new generative AI features that could solve that problem and others.
Not to be outdone, Zuckerberg says Meta is joining the race to build AGI. Unlike Google and OpenAI, Meta plans to make the AGI open-source once they figure out how to make it.
So, a free super-intelligent AI available to anyone to use as they see fit? Zuckerberg says it’ll be just fine. Are you sure, Mark?
True fakes or fake truths
The wave of AI fakes has blurred the lines between truth and fiction to the point where we’re struggling to decide what to believe. The rise of the “liar’s dividend” is a dangerous side effect of AI-generated misinformation.
When a message from a politician asks you not to vote for him, that should be your first clue that it’s a fake. A robocall impersonating Joe Biden was doing the rounds in New Hampshire telling voters not to bother voting in the primaries. As election season hots up we’re going to see a lot more of this.
Politicians and world leaders weighed in on generative AI at Davos where they were cautiously optimistic but did point to the risk of AI misinformation in politics.
This week the UK National Cyber Security Centre (NCSC) released a report warning of AI cyber threats. AI is going to make it easy for amateurs to give cybercrime a go and supercharge the efforts of the pros.
OpenAI goes back to school
ChatGPT Enterprise is making its first foray into educational institutions with OpenAI’s partnership with Arizona State University. They’re still trying to work out exactly what they’ll be using it for though.
In October, researchers at The University of Chicago unveiled their data poisoning tool called Nightshade. Version 1.0 is finally available to download and artists are using it to fight back against companies like OpenAI.
Stability AI also copped its fair share of criticism over its wholesale data grab to train its models. The company’s ex-VP of Audio launched a non-profit organization called “Fairly Trained”. It aims to certify companies that use “fair training data practices”.
The project is interesting, but will it make a difference in the fight over AI training data?
Speaking of fighting, OpenAI says its AI can now be used in military applications. The company quietly edited its Ts&Cs so that “military use” is no longer included in its list of prohibited uses. WarGPT anyone?
AI on the job
Will AI take your job anytime soon? An MIT study focused on the impact that computer vision will have on jobs offered a surprising answer to the question.
New York City became the first to pass a law regulating how employers use AI to automate the hiring process. The problem is, no one is checking up on them so very few companies are complying with the law.
There’s no way a human is wading through piles of resumes when an AI tool could pick the ideal employee in seconds. What elements of bias does this introduce into the hiring process though?
In a win for Team Human, a UK parcel firm had to disable its AI customer service bot after it went rogue. A disgruntled customer was able to get the bot to poetically slam the company when he couldn’t get through to a human service representative.
AI in the Orient
China is one of the bigger Asian players in the AI arena, but developments in other parts of the Far East are hotting up too.
Sakana, a Tokyo-based startup, raised $30m for its bio-inspired AI project. The company wants to create swarms of AIs inspired by birds, fish, or collaborative swarms of ants that could fundamentally change what AI will look like in the future.
In Korea, Samsung has teamed up with Google to bring Gemini to its AI-rich S24 series of mobile phones. The functionality that AI brings to Samsung’s new phones is impressive. But will people use it?
AI is a great listener
Ironically, humans are getting worse at listening to each other and having empathetic conversations as communication tech improves. A new study shows that lonely students are turning to AI chatbot companions to help with their mental health, with surprising results.
Diagnosing dementia can be a slow and difficult process. Researchers from the University of Sheffield developed an AI tool called CognoSpeak, which listens to a person’s speech to make a fast and accurate dementia diagnosis.
A lot of AI healthcare news has been about exciting research developments that may eventually help us in the future. We’re starting to see some exciting products become available right now. DermaSensor’s AI skin cancer detector has been approved by the FDA and will have an immediate effect on the lives of patients.
Should you be concerned about that mole or not? This device tells you in seconds.
In other news…
Here are some other clickworthy AI stories we enjoyed this week:
- An AI row at the Sundance Film Festival sparked a walkout.
- Sports Illustrated publisher laid off 100 employees after its AI snafu and more cuts may be coming.
- Alphabet cut its ties with the company that helped train Google’s AI products.
- OpenAI CEO Sam Altman is looking to raise billions to make AI chips.
- AI voice-generating company Eleven Labs raised $80m taking its value to $1b.
- SAP is restructuring 8,000 jobs as it shifts focus to AI.
- Ukraine will soon have more AI-powered drones in its military than soldiers.
Ukraine will have more drones than soldiers in its armed forces – what will such a dramatic change mean for the future of war? https://t.co/iSRe6STSIh
— New Scientist (@newscientist) January 25, 2024
And that’s a wrap.
Google blew me away this week with Lumiere. Let’s hope we get to play with it soon and that it’s as good as advertised. We still haven’t seen Gemini Ultra released yet so I’m not holding my breath.
Did the MIT report make you feel more job secure? I know they’re really smart at MIT, but I can’t help but think they seriously underestimated how many jobs AI is going to replace.
What do you think about Zuckerberg’s plan to release an open-source AGI in the wild one day? Great idea, or suicide by AI?
Let us know what you think and please send us links to any juicy AI stories we may have missed.
