AI News - May 2025 - part 2
[8] OpenAI ChatGPT's sycophancy problem
Blog post: "Expanding on what we missed with sycophancy". Apparently telling millions of users that their view of the world is insightful may be somewhat dangerous.
i strongly encourage everyone to read this blog post
— Aidan McLaughlin (@aidan_mclau) May 2, 2025
very detailed explanation of our posttraining, process, and what we’re changing to do better
link below pic.twitter.com/31kRubRq5w
[9] Interfacing your brain directly to the cloud
Ray Kurzweil explains how nanobots in your bloodstream will give you a direct link between your thoughts and a very powerful AI in the cloud.
Ray Kurzweil says we'll merge with AI non-invasively by the 2030s, using nanoscale sensors in the bloodstream that connect to the neocortex.
— vitrupo (@vitrupo) May 7, 2025
Your thoughts will sync with cloud-based intelligence.
“It will be as if it's in your mind.” pic.twitter.com/zA7Z4hcAn7
[10.1] The anti-AI movement will emerge suddenly
At the moment there is no organised effective movement against AI, but it takes only a single event. What will it be and what happens next?
Soon the anti-AI protest movement will be the largest movement in all of history,
— lethalintelligence.ai (@lethal_ai) April 30, 2025
but it will be too late... pic.twitter.com/fY10tfpvjo
[10.2] Best case, a Chernobyl size disaster
Stuart Russell says governments will allow AI companies to continue on. Maybe there will be a disaster big enough to cause controls to come in, or maybe we will not be that fortunate ...
Computer Scientist, Stuart Russell, predicts 4 future predictions:
— Haider. (@slow_developer) May 2, 2025
1) scaling up LLMs won’t lead to AGI
2) big AI labs already realize this and are exploring new methods, with AI likely to surpass humans within a decade
3) governments probably won’t act on AI safety until a… pic.twitter.com/bLhmnih4Zb
[11.1] AI with EQ (emotional quotient)
'capable of sculpting tone, style, and personality to make interactions feel personal and human' ...
Mustafa Suleyman describes the next generation of AI as a shift in relationship, not just functionality.
— Wes Roth (@WesRothMoney) May 2, 2025
These models aren’t just tools — they’re becoming emotionally intelligent, capable of sculpting tone, style, and personality to make interactions feel personal and human.… pic.twitter.com/FG5y71UoMm
[11.2] Facebook have AI Friends for you
Zuckerberg explaining how Meta is creating personalized AI friends to supplement your real ones: “The average American has 3 friends, but has demand for 15.” pic.twitter.com/Y9ClAqsbOA
— Roman Helmet Guy (@romanhelmetguy) April 30, 2025
[11.3] AI is already persuasive
BREAKING: The University of Zurich has been using AI bots to secretly manipulate Redditors since November 2024.
— Reddit Lies (@reddit_lies) April 28, 2025
The scariest part?
The bots were 6 times more likely to change the minds of Redditors than the baseline, often by leveraging misinformation.
More below🧵 pic.twitter.com/mZnhj6I9YS
[12] First publically available AI Scientist
So, what would you like it to make or discover?
Today, we are launching the first publicly available AI Scientist, via the FutureHouse Platform.
— Sam Rodriques (@SGRodriques) May 1, 2025
Our AI Scientist agents can perform a wide variety of scientific tasks better than humans. By chaining them together, we've already started to discover new biology really fast. With… pic.twitter.com/wMMmZoGZPI
[14] Image generation with Grok
The easy flow of image generation with Grok (also ChatGPT etc).
