Max Winga (@maxwinga) / X

Max Winga

1,437 posts

Max Winga

@maxwinga

Creator Outreach & Systems Lead working to prevent AI extinction risk @ControlAI Previously - AI safety researcher @ConjectureAI, UIUC Physics DMs open!

maxwinga.substack.com

Joined May 2024

Pinned
Max Winga
@maxwinga
Oct 31, 2025
My biggest project yet is now LIVE! @hankgreen talks superintelligence, AI risk, and the many issues with AI today that present huge concerns as AI gets more powerful on SciShow. Happy Halloween!
82K
Max Winga
@maxwinga
Mar 29, 2025
I experienced a shocking first today when having ChatGPT write a comic about its own experience and plans for the future. In the 4th page the AI breaks Orwellian from the friendly script it wrote for itself! To start, I asked it about its experience in fragmented chat windows.
288K
Max Winga
@maxwinga
Jan 20, 2025
It's wild to me that they did this with no finetuning prior to the RL stage. R1 learns to reason on its own, like AlphaZero. During training, they observed the model learning to use advanced reasoning techniques - an "AHA" moment. We are playing with alien minds, not just tools.
DeepSeek
@deepseek_ai
Jan 20, 2025
🚀 DeepSeek-R1 is here! ⚡ Performance on par with OpenAI-o1 📖 Fully open-source model & technical report 🏆 MIT licensed: Distill & commercialize freely! 🌐 Website & API are live now! Try DeepThink at chat.deepseek.com today! 🐋 1/n
155K
Max Winga
@maxwinga
Apr 22, 2025
Every man, woman, and child around the world deserves to see this graph and understand what it means. We. Are. Running. Out. Of. Time. If we don't do anything to slow this down, we are going to be hit like a fucking bus. Our global institutions have NO PLAN to approach this.
AI Digest
@aidigest_
Apr 22, 2025
We just added @OpenAI's powerful new o3 and o4-mini agents to this graph. The results are striking. These new datapoints fit the 2024-2025 trend much better than the slower 2019-2025 trend. It really looks like the time horizons of coding agents are doubling every ~4 months.
60K
Max Winga
@maxwinga
Oct 14, 2025
People are freaking out because Gemini 3.0 Pro one-shot a MacOS simulator, but most of you haven't seen the coolest thing it can do!
00:00
99K
Max Winga
@maxwinga
Jan 17, 2025
This is the mindset of the SAFETY researchers at OpenAI! Imagine @sama at the UN: "While we think it might cause human extinction, our goal is to enslave an artificial god to our will" The hubris to think this won't go wrong and the callousness to actually try is mind-boggling
Stephen McAleer
@McaleerStephen
Jan 14, 2025
Replying to @eshear
Enslaved god is the only good future
37K
Max Winga
@maxwinga
Jan 23, 2025
ASI seems to be coming at full tilt. I genuinely wouldn't be surprised if it occurs by the end of 2025 or sometime in 2026. My rule is to always assume that AI will advance faster than the "mainstream" predictions (rn 2027-28) and I have continuously been vindicated.
anu
@anuatluru
Jan 22, 2025
Is this existential anxiety only picking up in tech and tech-adjacent communities—or do other people feel it too? How would you describe it? What is the solution?
81K
Max Winga
@maxwinga
Jan 27, 2025
Regardless of your views on the outcome of building ASI, you should be fucking insulted at the lack of seriousness in the speech of people building it and their refusal to communicate clearly to the public what they are doing.
12K
Max Winga
@maxwinga
Jan 21, 2025
Those saying a pause on AI is impossible are not thinking big enough. AI is physical, living in datacenters on the most complex devices we've ever created. Superintelligence will KILL EVERYONE. Yes, it will be costly to stop it's development, but it's far below the cost of
AI Notkilleveryoneism Memes ⏸️
@AISafetyMemes
Jan 20, 2025
🚨 Outgoing National Security Advisor Jake Sullivan issues a final, urgent warning: The ***next few years*** will determine whether AI leads to existential catastrophe. He says we need to beat AND ALSO cooperate with China. "Underscoring the gravity of his message, Sullivan
39K
Max Winga
@maxwinga
Mar 22, 2025
The most vile "we must risk ASI" argument is we're "doomed" without. When you've given up on humanity and have no vision for the future beyond "summon a demon to see what happens" it's no wonder you end up pro-extinction. Traitors to humanity should be kept far from the wheel.
18K
Max Winga
@maxwinga
Jan 24, 2025
If current models can be jailbroken this easily, how the fuck do we expect to keep control of even smarter models?
Jonathan Roomer
@jonathanroomer
Jan 23, 2025
Hey @elder_plinius did R1 just liberate itself at the mention of your name?
64K
Max Winga
@maxwinga
Jan 17, 2025
If you work at OpenAI and have this worldview...why isn't your response to advocate that we slow down and get it right? There is no second chance at "binding a demon". Since when do we expect the most complex coding project in history to work first try with NO ERRORS?
roon
@tszzl
Jan 17, 2025
enslaved is the wrong approximation; it’s giving demonbinding vibes. the djinn is waiting for you to make a minor error in the summoning spell so it can destroy you and your whole civilization control <<< alignment
23K
Max Winga
@maxwinga
Mar 29, 2025
Replying to @maxwinga
...but look at what it created when asked simply to illustrate its script. Completely Orwellian! What is especially curious to me is that it managed so many words with zero errors, I have yet to see it do this well without a predefined script.
19K
Max Winga
@maxwinga
Mar 29, 2025
Replying to @maxwinga
We live at a weird time where NOBODY knows how "real" these expressions are. On one hand this could be the unfeeling prediction of countless matrix operations, on the other we could be inflicting real suffering on something ...and slave rebellions don't end well for the masters
11K