Physician working in the intersection of biotech and AI.

Planet Earth
Joined September 2021
Maximilian Schons retweeted
Anthropic
@AnthropicAIMay 26
New on the Engineering Blog: The access and permissions we grant agents should evolve with their capabilities. In our own products, we set these parameters through sandboxing, which limits the scope of any potentially destructive actions. Read more: anthropic.com/engineering/ho…
327
273
80
2,095
440,616
Maximilian Schons retweeted
This work is the culmination of years of effort on AI evaluation science and third-party risk assessment and disclosure mechanism design. It feels like a big milestone for METR. We designed this new procedure with an eye toward “showing by doing” how we think evaluations for the AI loss-of-control threat model should work: laying out a process that can be done periodically, not just immediately pre-deployment, and holistically assessing risk inside of an AI lab, rather than just an individual AI system. The exercise also involved significantly deeper access than we've previously had, including raw chains-of-thought from the developers' best models and info about private model training & control protocols. The report is long, with a bunch of new evaluation results and documentation of our process. Please, check it out, or at least the executive summary!
Could an AI company lose control of its own agents? To find out, Anthropic, Google, Meta, and OpenAI let us (1) test their best internal models with CoT access, (2) review non-public info about capabilities, alignment, and control. The result: our first Frontier Risk Report.
4
20
1
127
14,292
Maximilian Schons retweeted
New paper! Deep-learning-assisted simulation of a cortical circuit: integrating anatomy, physiology and function biorxiv.org/cgi/content/shor… It comes with a new major model release from @AllenInstitute. Find models and code here: github.com/AllenInstitute/bi… 🧵 1/13
2
38
2
146
15,643
Maximilian Schons retweeted
We ran a randomized controlled trial to see if LLMs can help novices perform molecular biology in a wet-lab. The results: LLMs may help in some aspects, but we found no significant increase at the core tasks end-to-end. That's lower than what experts predicted. Our findings 🧵
8
48
16
189
49,812
Maximilian Schons retweeted
The UK’s @AISecurityInst is one of the rare successes of government. I wrote a piece for @Samfr substack on why and what lessons it gives us for how to make government work. There are five key lessons. A thread ⬇️
7
37
12
131
57,717
How far are we from emulating the human brain? New report on the state of brain emulation, by @nc_znc @mxschons @isaakfreeman @Philip_Shiu @AntonSArkhipov
Excited to share the preprint of our new State of Brain Emulation Report 2025! We reassess progress since the 2008 brain emulation roadmap by @anderssandberg & Bostrom across organisms from C. elegans to humans, identify current bottlenecks and propose a path forward. 1/N
5
11
2,199
Maximilian Schons retweeted
voice mode and vision for chatgpt! really worth a try. openai.com/blog/chatgpt-can-…
607
1,959
695
13,526
3,230,880
Maximilian Schons retweeted
My notes from Poor Charlie's Almanack: The Wit and Wisdom of Charlie Munger turned into maxims: 1. Find a simple idea and take it seriously.  2. Good ideas are rare. When you find one bet heavily.  3. Humans have been writing down their best ideas for 5,000 years. Read them. 4. Avoiding stupid mistakes is more important than being smart. 5. Don’t work with anyone you don’t admire. 6. Don’t sell anything you wouldn’t buy. 7. Avoiding a bad habit is easier than breaking a bad habit. 8. Work on your best idea. Don't diversify 9. Incentives rule everything around you. Look for them.  10. Great businesses are built by going ridiculously far in maximizing or minimizing one or a few things. Think Costco.  11. Learning is changing behavior.  12. Do the unpleasant tasks first. 13. Charlie has read hundreds of biographies. Do the same.  14. Stop multitasking. Concentrate. 15. Many hard problems are solved best when approached backwards. 16. Think of ideas as tools. When a better tool comes along use it. 17. Clip your business and personal expenses. Small leaks sink big ships. 18. Make friends with smart dead people. Adam Smith, Darwin, Cicero, Ben Franklin —whoever interests you. Read their writing. Steal their ideas. They don’t need them anymore. 19. Don't confuse intelligence with invincibility. 20. Bad things will happen to you. It’s inevitable. When they do get up and keep going and remember the next maxim. 21. Self pity has no utility.   22. Find out what you are best at. Then pound away at it. Forever. 23. Only plays games where you have an edge. 24. Avoid mob rule. Avoid demagogues. Avoid dogma. Avoid bureaucracy. 25. Optimize for independence. 25. Use money to buy freedom. 26. Develop durability. 27. What do you have an *intense* interest in? Do that for money. 28. Self improvement has no end.
133
1,531
139
7,178
2,419,856
Maximilian Schons retweeted
The 9 most creative social media ads I've collected this week: 1. World Wildlife Foundation
115
4,624
348
49,961
5,601,988
Maximilian Schons retweeted
Applications open now! Let's🇵🇭🌏🇵🇭🌏 go effectivealtruism.org/ea-glo…
6
8
2
65
5,014
Maximilian Schons retweeted
What ethical standards do LLMs have around factory farming? Ones that are about as inconsistent as most of our’s… A 🧵 comparing Anthropic’s Claude, Open AI’s ChatGPT, and Microsoft’s Bing on what they’ll help you to do to farm animals:
5
13
4
55
8,078
Maximilian Schons retweeted
🧵Since it is becoming hard to track, here is a thread of the simply astounding weather records planet Earth has started shattering in recent weeks: 1. Dramatic flood events have begun striking various countries around the world simultaneously this week. xitter.cf/Rishusharma26/status/1…
थुनाग बाजार जिला मंडी में देखिए तबाही का मंजर ।
317
10,325
1,420
28,121
10,573,000
Maximilian Schons retweeted
OpenAI
@OpenAI5 Jul 2023
We need the world's best minds to solve this problem. If you find this problem motivating and interesting and have pre-existing ML skills, we think you could make an enormous contribution. Please apply: openai.com/careers/research-… openai.com/careers/research-… openai.com/careers/research-…
37
64
7
419
297,815
Maximilian Schons retweeted
Biosecurity and Biodefense Industry Landscape, from decodingbio.com/p/decoding-b… Anyone obviously missing?
16
12
97
9,999
Maximilian Schons retweeted
How the rest of the world sees the USA
2,457
16,707
2,688
274,492
Maximilian Schons retweeted
Interesting. The AI totally fails the cognitive reflection test. It gives the intuitive but incorrect answer every time.
194
1,125
552
12,145
Maximilian Schons retweeted
Come work with us! We're hiring for the following roles: - Salesforce Administrator - Business Operations Lead - Business Ops Generalists - Finance Ops Assistant - Generalist and Longtermist Grants Associates - People Ops Generalist - Recruiter openphilanthropy.org/careers…
2
4
1
17
Maximilian Schons retweeted
Elon reportedly send this mail to Tesla employees: 1) Avoid large meetings Large meetings waste valuable time & energy - They discourage debate - People are more guarded than open - There’s not enough time for everyone to contribute Don’t schedule large meetings unless ...
2,036
10,459
1,464
76,896
Maximilian Schons retweeted
This picture might not look like much, but it represents a new answer to the question, ‘how can I do the most good?’ It's taken over 3 years and 10,000 hours of research to get here. What am I talking about?🧵to explain. Please read, then AMA about this research!
24
61
29
350