@aiclerk Profile picture

John Lamb

@aiclerk

Fortune 100 in-house counsel here to learn about AI. RTs and likes ≠ endorsements. Tweets are my own.

Joined February 2023
Similar User
simplexico photo

@_simplexico

TypeLaw photo

@TypeLaw

David Murray photo

@DavidMGVA

Freya Zhou photo

@freya_zhou

John Lamb Reposted

And modify. And combine. And reject. And use the AI as inspiration. And know when to avoid using it for inspiration. To be clear, working with AI well is not a passive process, but it also isn't limited to the patience of a human editor or co-author.


John Lamb Reposted

We are just not used to abundant "intelligence" (of a sort), which leads people to miss a huge value of AI. Don't ask for an idea, ask for 30. Don't ask for a suggestion on how to end a sentence, ask for 20 in different styles. Don't ask for advice, ask for many strategies. Pick


John Lamb Reposted

The danger is that AIs can seem really smart and be wrong. Without benchmarks or the ability for most people to see whether the AI is great or faking it, we will not have a good intuition for how useful it is.


John Lamb Reposted

A gap in the work of the AI labs is that they don’t seem to see the next phase of AI as about management: how do we verify & control the work of flawed semi-autonomous AI agents. Current UX designs for agents are all about autonomy. We need to focus on feedback loops & check-ins


John Lamb Reposted

So far, I have not seen a single approach to integrating documents into LLMs that completely prevents the AI from hallucinating about document contents. With some approaches, hallucinations are relatively rare, but they happen. Use cases need to take hallucinations into account.


John Lamb Reposted

A good illustration of the Jagged Frontier, where the relative difficulty of tasks to humans is not helpful in understanding how easy tasks are for AI It does chess but not tic-tac-toe. It writes sonnets but not 50 word paragraphs. It passes the bar exam but gets quotes wrong.

Oddly, gpt-3.5-turbo-instruct still cannot play tic-tac-toe. I tried many prompts, with and without board state, few-shot, etc. This is cherry-picked as the *best* game. If you're skeptical, go try your best prompts and report back. This remains the biggest mystery in AI, imo.

Tweet Image 1


John Lamb Reposted

People are making using AI for work too hard. Just use it wherever you ethically & legally can. Don't delay by fine tuning, or training on your data, or building a customized solution. Just get access to whatever the cutting edge model is & try it, you can do other stuff later.


“A baby spitballing credible Shakespeare imitations” is not an AI analogy I had on my Bingo card I can see that. With both, parents say, “If you think that’s impressive, wait till you see what the kid cooks up tomorrow”

Oddly, gpt-3.5-turbo-instruct still cannot play tic-tac-toe. I tried many prompts, with and without board state, few-shot, etc. This is cherry-picked as the *best* game. If you're skeptical, go try your best prompts and report back. This remains the biggest mystery in AI, imo.

Tweet Image 1


Yep. Professionals handling confidential data will be looking for security commitments from LLM #AI tools. More flexibility when processing publicly available information. #lawtwitter #legaltech

Full thread with >1800 comments (!) on the GPT-4 release: news.ycombinator.com/item?id=351545… Here’s a rebuttal to the post above:

Tweet Image 1


Where’s the #ChatGPT-like tool for my Twitter activity? Ideally would 1️⃣ read and summarize all the #legaltech #AI tweets I’ve liked or retweeted, including linked pages, and 2️⃣ synthesize into top takeaways. Next-gen tools may do this. Piecemeal options for now. #lawtwitter


John Lamb Reposted

A recent study asked professionals to write realistic memos, strategy documents and policies. The ones who were given ChatGPT completed tasks 37% faster, and their average writing quality increased as well. All of this is without added training or extensive experience using…


John Lamb Reposted

“It is entirely possible to be cautious about using ChatGPT in legal professional environments (because it has not been engineered for that use) and simultaneously excited about the potential use of generative AI in the legal industry.” - @Nicola_Shaver


John Lamb Reposted

My thought is they should have named ChatGPT “First Drafts.” If you know how to do something, ChatGPT lets you do it faster and sometimes better.

ChatGPT is a powerful productivity tool. But most people suck at using it effectively. 6 ways to make ChatGPT your productivity machine in 2023:



An 80/20 rule for #AI accuracy in #legaltech #lawtwitter

Legal engineers: if we are going to apply ChatGPT to the legal system we need to focus -- for awhile at least -- on applications where an 80% reliable answer is good enough.



John Lamb Reposted

This doesn’t mean these jobs are going away, and they may change in ways good (less tedious work) or bad (fewer jobs). This “could involve substitution or augmentation depending on various factors associated with the occupation itself.” But these jobs should expect to see change


“Legal services” at top of list on right #legaltech #lawtwitter #ai

Here it is, everyone, the first good model of which industries & jobs are most at risk of disruption from the sudden boom in generative AI, using an established method that looks at job tasks. If your job or industry is on these lists, you should be paying close attention to AI.

Tweet Image 1
Tweet Image 2


John Lamb Reposted

I keep finding things that AI can do that are unexpected, even knowing how LLMs work, from applying philosophical theories in novel ways to coming up with a plan for saving the Roman Empire. I compiled a bunch of interesting examples in this post. oneusefulthing.substack.com/p/feats-to-ast…


John Lamb Reposted

👇Terrific GPT observation from math superstar @TerenceTao6 Peer review and comprehension just got harder.

Tweet Image 1

John Lamb Reposted

AI hallucinations can be more convincing than the original. I asked Bing to read the 1885 book "100 Proofs the Earth is Not a Globe" & give examples of rhetorical techniques. It gave awesome quotes of flat earth arguments that seemed very 1885, but none of them were in the book!

Tweet Image 1

United States Trends
Loading...

Something went wrong.


Something went wrong.