
Biologist & Nutritionist 🧬 | AI Enthusiast 🤖 | Exploring the intersection of biology and AI. Sharing insights on #AI, #Biotech, #Futureofwork
Italia
Joined November 2023
- Tweets4.3K
- Following638
- Followers762
- Likes48.7K
Pinned Tweet
Replying to @emollick
Expanding on point 2:
In my profession, and discussing with experts from other fields, I’ve noticed that today’s top AI models work best when guided by domain specialists.
For example, someone without expertise in fields like mine (nutrition) or medicine might miss subtle nuances that make an AI's response subtly incorrect.
Non-experts often struggle to ask precise questions that elicit truly professional answers from AI.
That’s why, at this stage, I see current AI models as ideal "coworkers" for experts rather than standalone solutions. We need human expertise to unlock their full potential and avoid critical mistakes.
This is something I've been hoping for: near GPT-6 Astra-level performance in Ultrafast mode, without eating into my usage limits.
GPT-6.2 Sol with ULTRAFAST could be incredibly useful for some of my tasks.
1
9
402
Agreed. We should probably just forget about it, and maybe it'll finally drop a day before the end of the year. Unless Anthropic releases Fable 5.5/Opus 5.6 (which seems likely) and OpenAI decides to scrap Astra 6.2 altogether.
I guess we'll be stuck at Astra 6-level capabilities for quite a while.
Our only hope of keeping the iterative deployment of AI models going is for China to keep pushing relentlessly. The West needs serious competition from other players to keep AI development moving forward before superintelligence gets banned.
I think it would be a disgrace if we stalled at the current level of AI capabilities. Frontier AI models are already somewhat useful, but they're still far from being able to truly transform how we work and accelerate human progress. That's why I sincerely hope other countries keep pushing, too.
2
3
166
One thing I think many people get wrong is assuming that if AI labs reach RSI, model releases (or public deployments) will become more frequent. Not necessarily.
In fact, I suspect that as we approach RSI, releases could become even less frequent because of safety concerns, competitive moats, and political pressure, including the frankly alarming proposals from some Dems to pause advanced AI development or even ban superintelligence outright.
Internal progress could accelerate dramatically while public access slows to a crawl.
1
7
370
What I hate most is the complete silence around upcoming AI releases. Why not at least try to be transparent and give rough release windows?
Something like: "We're aiming to release model X in the second half of November, assuming everything goes well."
Then, if it doesn't happen, just explain to your paying customers what went wrong or why things changed.
Is the competition really so fierce that AI labs would rather piss off their paying customers than give competitors a glimpse of their roadmaps?
GPT-6.2 is like a diet: it's always starting next Monday, but somehow that Monday never comes
I think that a lot of paying OpenAI subscribers (and I'm one of those) don't want more features or gimmicks right now. They mainly want a new, more capable and reliable model, be it Sol or Astra.
Especially now that Fable 5.2 is out, OpenAI really needs to ship, or at least be more open about why there's still no sign of the next model.
I think users would really appreciate some transparency.
@OpenAI
Between his recent interview on AI and the military and this video, I'm starting to like Dario again.
Maybe he's not a doomer after all, or at least not the toxic kind. We might still be on the good timeline... 🤞🏻
Anthropic CEO predicts cancer, Alzheimer's, heart disease, mental illness will be cured in less than 10 years
ACCELERATE
12
1,336
Imagine a backlash so strong that we end up stuck for years with today's publicly available models: pretty good AI, but still far from being truly transformative.
I use GPT-6 Astra every day, and I can tell you that such a future would be painful. These models are already very useful, but still imprecise and "AI-ish" enough to require constant checking, corrections and refinement.
Now, imagine knowing that much more capable systems EXIST and are used behind closed doors, but being told: "Sorry, you can't have them. We're keeping them gated until we're absolutely sure NOTHING can ever go wrong." As if humans weren't perfectly capable of screwing things up in the meantime.
I can understand the argument against accelerating too fast, but we also need to make sure we don't end up at the opposite extreme: a prolonged painful pause.
If capabilities keep improving and being deployed at a reasonable pace, we can keep figuring things out along the way. But getting stuck at this awkward halfway point for years? That would be dreadful.
Everything but frontier models, apparently.
Hope I'm wrong, but I don't expect any new frontier model releases or even announcements this week.
We're turning our lessons from defending OpenAI into practical tools for our customers. On Thursday, join us at Cyber at the Frontier, a live keynote, for our latest cyber announcements.
Working with our frontier models, we've identified and resolved important vulnerabilities in our own code and infrastructure. That experience is informing the tools and playbooks we're building to help other security teams strengthen their defenses.
@gdb shares what we've learned and why security teams should keep putting the latest models to work as capabilities improve.
Tune in on November 12 at 10 a.m. PT: openai.com/cyber-at-the-fron…
Nice to see Google pushing hard. We're now living in a timeline where Google ships frontier models faster than @OpenAI 💀
Gemini 4 Carbon, now available in the Gemini API and to Ultra subscribers!
Just a few weeks after Argon was launched, this model makes significant progress on lots of the feedback from Argon and pushes the frontier in cyber, knowledge work, and code! It's a great model : )
1
32
984


















