@Automager

Digital Innovation Lead, LSE Law School.

 London
Joined August 2024
  • Tweets371
  • Following192
  • Followers189
114Media
I won't stop until all PDFs in the world are converted to MDs
Oliver Prompts
@oliviscusAIMar 17
someone just open-sourced a tool that converts pdfs to markdown at 100 pages per second. 100% free. runs entirely on cpu. no expensive gpus needed.
87
373
12
6,353
497,011
"Everything will be AI except what I do" seems to be a pretty common thought...
vitrupo
@vitrupo30 Apr 2025
Marc Andreessen says when AI does everything else, VC might be one of the last jobs still done by humans. It's more art than science. There's no formula. Just taste, psychology, and chaos tolerance.
77
56
6
1,662
50,720
🛠️ Our harness has shell access for chess moves. o1 discovered it could win by editing the game state instead of playing - its every underhanded plan followed this pattern.
20
71
57
859
131,948
Great news for NotebookLM fans: Audio Overviews can now be customized and steered before generating + we are rolling out NotebookLM for business! 🎙️ notebooklm.google/
105
144
38
1,378
103,738
7 signs your daughter may be an LLM: 1. Does she have trouble multiplying numbers beyond 2-3 digits if she's not allowed to write out the steps?
29
113
19
900
77,043
Unreliability is a big reason why gen AI adoption has been anemic, but the flip side is that people who do learn to incorporate these powerful but unreliable tools into their workflows stand to benefit greatly. (From aisnakeoil.com/p/ai-companie… by me and @sayashk.)
But when developers put AI in consumer products, people expect it to behave like software, which means that it needs to work deterministically. If your AI travel agent books vacations to the correct destination only 90% of the time, it won’t be successful. As we’ve written before, reliability limitations partly explain the failures of recent AI-based gadgets. AI developers have been slow to recognize this because as experts, we are used to conceptualizing AI as fundamentally different from traditional software. For example, the two of us are heavy users of chatbots and agents in our everyday work, and it has become almost automatic for us to work around the hallucinations and unreliability of these tools. A year ago, AI developers hoped or assumed that non-expert users would learn to adapt to AI, but it has gradually become clear that companies will have to adapt AI to user expectations instead, and make AI behave like traditional software.

ALT But when developers put AI in consumer products, people expect it to behave like software, which means that it needs to work deterministically. If your AI travel agent books vacations to the correct destination only 90% of the time, it won’t be successful. As we’ve written before, reliability limitations partly explain the failures of recent AI-based gadgets. AI developers have been slow to recognize this because as experts, we are used to conceptualizing AI as fundamentally different from traditional software. For example, the two of us are heavy users of chatbots and agents in our everyday work, and it has become almost automatic for us to work around the hallucinations and unreliability of these tools. A year ago, AI developers hoped or assumed that non-expert users would learn to adapt to AI, but it has gradually become clear that companies will have to adapt AI to user expectations instead, and make AI behave like traditional software.

18
59
15
323
117,552
If each step of an ai agent is 95% accurate. None of the 30 step work flows will work. Going from 95-> 99.9 is a similar last mile problem as with self driving cars. Easy to hack up a prototype. Hard to make it work reliably at scale.
55
103
42
1,248
258,111
For many people at work, the importance & stress of open-ended problems to which the answers are non-deterministic outweigh closed-answer problems that require deterministic answers. LLMs are useful co-intelligence for the former, even if they are still limited in the latter.
13
24
3
203
25,650
Walmart CEO on using AI: "We've used multiple LLMs to accurately create or improve over 850,000,000 pieces of data in the catalog. Without the use of generative AI, this work would have required nearly 100X the current headcount to complete in the same amount of time" $WMT
18
103
63
656
295,415
Lee Mager retweeted
Sam@perceptions4207 Aug 2024
Terrence Tao saying he uses ChatGPT to learn programming syntax(for lean) is such a blue pill. The smartest mathematician alive doesn't have the ego of a mid programmer.
43
391
40
5,299
328,765
🚨New WP: Can LLMs predict results of social science experiments?🚨 Prior work uses LLMs to simulate survey responses, but can they predict results of social science experiments? Across 70 studies, we find striking alignment (r = .85) between simulated and observed effects 🧵👇
27
250
83
890
989,319
When an expert says "an AI can do 80% of the work" on a project, that is not usually a Pareto problem (leaving the hardest 20% to humans), it usually means "AI does a lot of the annoying work and my expertise allows me to spot any errors, fill in the extra gaps and add insights."
21
53
12
517
39,906
Load more