Welcome to this week’s roundup of AI information made by people, for people.
This week, OpenAI advised us that it’s fairly certain o1 is kinda secure.
Microsoft gave Copilot a giant enhance.
And a chatbot can treatment your perception in conspiracy theories.
Let’s dig in.
It’s fairly secure
We had been caught up within the pleasure of OpenAI’s launch of its o1 fashions final week till we learn the positive print. The mannequin’s system card presents attention-grabbing perception into the security testing OpenAI did and the outcomes could increase some eyebrows.
It seems that o1 is smarter but in addition extra misleading with a “medium” hazard degree in keeping with OpenAI’s ranking system.
Regardless of o1 being very sneaky throughout testing, OpenAI and its purple teamers say they’re pretty certain it’s secure sufficient to launch. Not so secure in case you’re a programmer on the lookout for a job.
If OpenAI‘s o1 can cross OpenAI‘s analysis engineer hiring interview for coding — 90% to 100% charge…
……then why would they proceed to rent precise human engineers for this place?
Each firm is about to ask this query. pic.twitter.com/NIIn80AW6f
— Benjamin De Kraker 🏴☠️ (@BenjaminDEKR) September 12, 2024
Copilot upgrades
Microsoft unleashed Copilot “Wave 2” which can give your productiveness and content material manufacturing an extra AI enhance. Should you had been on the fence over Copilot’s usefulness these new options would be the clincher.
The Pages characteristic and the brand new Excel integrations are actually cool. The best way Copilot accesses your knowledge does increase some privateness questions although.
Extra strawberries
If all of the latest discuss OpenAI’s Strawberry venture gave you a yearning for the berry you then’re in luck.
Researchers have developed an AI system that guarantees to remodel how we develop strawberries and different agricultural merchandise.
This open-source utility might have a huge effect on meals waste, harvest yields, and even the value you pay for contemporary fruit and veg on the retailer.
Too straightforward
AI fashions are getting so sensible now that our benchmarks to measure them are nearly out of date. Scale AI and CAIS launched a venture known as Humanity’s Final Examination to repair this.
They need you to submit robust questions that you just suppose might stump main AI fashions. If an AI can reply PhD-level questions then we’ll get a way of how shut we’re to attaining expert-level AI programs.
Should you suppose you might have a superb one you would win a share of $500,000. It’ll need to be actually robust although.
Curing conspiracies
I really like a superb conspiracy idea, however among the issues folks consider are simply loopy. Have you ever tried convincing a flat-earther with easy info and reasoning? It doesn’t work. However what if we let an AI chatbot have a go?
Researchers constructed a chatbot utilizing GPT-4 Turbo and so they had spectacular ends in altering folks’s minds in regards to the conspiracy theories they believed in.
It does increase some awkward questions on how persuasive AI fashions are and who decides what ‘truth’ is.
Simply since you’re paranoid, doesn’t imply they’re not after you.
Keep cool
Is having your physique cryogenically frozen a part of your backup plan? If that’s the case, you’ll be pleased to listen to AI is making this loopy thought barely extra believable.
An organization known as Choose AI used AI to speed up the invention of cryoprotectant compounds. These compounds cease natural matter from turning into crystals throughout the freezing course of.
For now, the applying is for higher transport and storage of blood or temperature-sensitive medicines. But when AI helps them discover a actually good cryoprotectant, cryogenic preservation of people might go from a moneymaking racket to a believable choice.
AI is contributing to the medical discipline in different ways in which may make you a bit nervous. New analysis exhibits {that a} stunning quantity of docs are turning to ChatGPT for assist to diagnose sufferers. Is {that a} good factor?
Should you’re enthusiastic about what’s occurring in medication and contemplating a profession as a physician it’s possible you’ll need to rethink that in keeping with this professor.
That is the ultimate warning for these contemplating careers as physicians: AI is turning into so superior that the demand for human docs will considerably lower, particularly in roles involving normal diagnostics and routine therapies, which shall be more and more changed by AI.… pic.twitter.com/VJqE6rvkG0
— Derya Unutmaz, MD (@DeryaTR_) September 13, 2024
In different information…
Listed here are another clickworthy AI tales we loved this week:
Gen-3 Alpha Video to Video is now accessible on net for all paid plans. Video to Video represents a brand new management mechanism for exact motion, expressiveness and intent inside generations. To make use of Video to Video, merely add your enter video, immediate in any aesthetic course… pic.twitter.com/ZjRwVPyqem
— Runway (@runwayml) September 13, 2024
And that’s a wrap.
It’s not stunning that AI fashions like o1 current extra danger as they get smarter, however the sneakiness throughout testing was bizarre. Do you suppose OpenAI will keep on with its self-imposed security degree restrictions?
The Humanity’s Final Examination venture was an eye-opener. People are struggling to search out questions robust sufficient for AI to unravel. What occurs after that?
Should you consider in conspiracy theories, do you suppose an AI chatbot might change your thoughts? Amazon Echo is all the time listening, the federal government makes use of large tech to spy on us, and Mark Zuckerberg is a robotic. Show me fallacious.
Tell us what you suppose, comply with us on X, and ship us hyperlinks to chill AI stuff we could have missed.
