Dan Marcus

All in all, a very mixed bag. For all the talk of Open AI cannibalising those building on top of their platform, there is still plenty of room for companies with domain knowledge to beat the general functionality Open AI is offering up. At least, until the next update!

If you give it the right information and closely track relevant wording in the regulation it does well and can even provide answers that require looking at two different sections (the definition of CASS medium firm isn’t close to the requirement to allocate a director to a CMAR)

I decided to give ChatGPT 4’s new document reading a little test with some regulatory querying. Here’s what I learnt (a 🧵)

…but don’t ask for the cross reference as it will give you the wrong one 🤦🏻‍♂️

Tell it to focus on that right track and it can give a good answer…

Give it a slightly harder question and it can give you an answer to another question, before giving you a hint of the real answer

@mikeandallie If you need to find out the changes between the documents and have copies of both, I’d be happy to run a comparison for you. Just DM me

2/2 The pic in the above tweet shows a paragraph from Chapter 1. The sentence highlighted in blue was in the version approved in June, and it was removed in the October version. I'd love to know more about why this sentence was deleted.

1/2 It took about a week to read what I thought was the final draft of the California Math Framework back in June. I wish a document showing the changes from the June version to the October version would have been published. Some of the changes are meaningful, and some aren't.

Here, in full directly on Twitter, is "A Hackers' Guide to Language Models". This 90 minute tutorial is designed to be the one place I point coders at when they ask "hey, tell me everything I need to know about LLMs!" It covers both @OpenAI models and open source ones in depth.

@oliver_twiste People who don’t want to shower for 3 days…

@lawheroezV2 @legalytical It’s white-labelled, but I think it’s litera nder the hood. Knowing my firm they’re hosting it on internal servers, but it’s still better than running them locally!

@ThisIsDanMarcus @legalytical Interesting and that’s why I think the question of value (solving a real pain) is more dominant in relation to adoption. Which tool is this?

My slight brain dissonance:we talk a lot about users in legal needing “intuitive””easy to use” products to remove friction/minimize behavior change and yet MS Word,the most highly adopted Legaltech tool in the planet,is as unintuitive as it gets+ gives users big brain freezes🤯

@lawheroezV2 @legalytical Another observation - my firm recently launched a cloud doc comparison tool. It is buggy in a way that requires some specific habits, but I can now compare 10+ docs direct from IManage simultaneously using cloud computing (faster, doesn’t kill my v limited RAM), so I will learn!

@legalytical Hence the dissonance. Is it really a learning issue ?

@lawheroezV2 @legalytical I think the answer to this is inertia. MS Word gets away with this because it is the default. It takes a lot to move away from default behaviours, so “ease of use” becomes important to lower the burden

@jkubicki Then still has to hit 2000 hours billable…

Traditional lawyer: writes a blog and ignores social media Modern lawyer: has a newsletter and teaches, shares, learns and grows audience (potential clients, referrals, partners) via social media

Lots of people talking about cost of nursery fees. We’re grappling with this at the moment. There’s one nearby that cost £71 for a HALF day. Full time £113.60. That’s £1,420 for 4wks half days £2,272 for 4wks full time. Approx £18,460 part time annually. Or £29,536 full time

@ZachAbramowitz Any examples?

Some of the most expensive legal services are about to become MUCH cheaper. Some of the most lucrative legal work is about to become a LOT less profitable. The timeline is going to shock us.

@alexgsmith I’d happily discuss search with you - I have no buying power though!

@alexgsmith I doubt biglaw would allow an integration like this to be turned on. To the extent it is sending it to a cloud hosted instance of GPT, the context would be too sensitive. Biglaw tech isn’t given a lot of credit, but they know the secret sauce is in the firm data…

Ok what

This chart is amazing. After the calamity of Trussonomics, expected BoE interest rates soon fell back. Equivalent mortgage rates shot up like a rocket but have fallen like a feather …leaving households still at the sharp end of the moron premium