AI Reading for Friday February 28
GPT 4.5, Grok as based MAGA "You're fired" AI
OpenAI announces GPT-4.5, initially warns it’s not a frontier AI model - The Verge
I mean, it’s an order of magnitude bigger but I guess only reasoning models count as frontier models now?
Super expensive, worse than o3-mini and o1 but costs more. I ran today’s newsbot on it and I’m poor now. (well, cost $10 instead of like $1-$1.50). Obv will be rate-limited in the Plus subscription, maybe even Pro.
Seems like a bragging rights exercise at this price, get something in LMarena that maybe beats Grok-3 and Gemini. Also they want to collect data on how people might use it. But where is GPT-4.5-mini, couldn't they have distilled 4.5 into something a little better than 4o but same price?
If it doesn’t beat Sonnet 3.7 which was trained for a lot less money it’s kind of a dud. It will right? And they will come out with a cheaper version soon right?
o3-mini is good but Sonnet and DeepSeek so hot right now.
Industry observers a bit underwhelmed by 4.5, especially the price point. - VentureBeat
Mark Chen says scaling isn't dead, 4.5 is just the next evolution, hiccups along the way to releasing it are par for the course, making it more efficient will come soon. - Alex Kantrowitz, Big Tech
4.5 said more human-like, more factual / less prone to hallucinate - Business Insider
Musk, DOGE, and the AI-Fueled Plot to Fire Everybody - Intelligencer
Internal documents show how Grok is trained to be the based MAGA chatbot - Business Insider
Interesting examples at the bottom. IMHO if the majority group is biased against the 10% minority group, the majority group encounters bias 10% of the time, the minority 90% of the time. Obv everybody should try to be less racist but also obv there is disparate impact and a disparate history. And like a virus, a small change in transmission makes conditions endemic or not endemic. I know it’s woke to believe in systemic anything and only individual responsibility, and that’s why we can’t have nice things, people don’t believe in common decency or responsibility for society as a whole any more. Might makes right!
Anthropic’s C.E.O., Dario Amodei, on Surviving the A.I. Endgame - NY Times
The Coming AI Down Rounds; Plus, Always Be Raising - The Information
Big Tech wants everyone to use AI, just not when applying to their jobs - Business Insider
Bringing AI Agents to Healthcare - WSJ
These humanoid robots could build themselves on factory lines - Business Insider
Figure says it will field test humanoid robots for household chores this year - New Atlas
A self-taught AI robot can already bunny hop a bike better than me - Singletracks Mountain Bike News
Teleperformance uses AI to soften call center workers' accents in real time - Yahoo
Chatbots can teach humans to be better at empathy - WSJ
Microsoft releases Copilot for Mac - The Verge
Robot delivery vehicles on the loose - Long Beach Post News
Do I need to be worried about getting run over by these things driving on the sidewalk doing DoorDash?
Some recent AI EconTalk podcasts - EconTalk
Most of those guys sound super annoying but this was a good kicker quote:
Things that are vastly more intelligent than you are really hard to understand and predict; and the wildlife next door, as much as we might like it, we will also build a parking lot over it at a heartbeat and they’ll never know why. They’ll never know why. It’s totally beyond their ken. So, when you live on a planet next to things that are far vastly smarter than you or anyone else, they are the humans in that scenario. They might just build a parking lot over us, and we will never, ever know why.
Pinterest Is Being Strangled by AI Slop - Futurism
Is AI the Key to Solving the World's Toughest Math Problems? - IEEE Spectrum
AI helps interpret full-body MRIs - Health Care IT News
Fake jellyfish video goes viral - Snopes
No word on whether this one is real - Reddit
Me and my monster - Reddit
Did you hear... - YouTube
Follow the latest AI headlines via SkynetAndChill.com on Bluesky