'Building AI co-workers going to be largest opportunity of tech in our lifetime' Jake Heller, head of product for CoCounsel – an AI bot from multinational information conglomerate Thomson Reuters – has a tip for anyone selling AI software: Don't sell software. Rather, he argues, sell AI as a co-worker. Corporations understand the cost of people, but not code, he reasons – and it has worked for him at least …
He gives legal work as an example, but LLM has already failed in legal work.
He also said "Even if it's truly net positive in the aggregate, I think some individuals in some job categories and some tasks are gonna be harder hit than others and faster than others. And so, we're going to rise to the challenge of massive potential for displacement, shifts, et cetera."
"Some of you may die, but it's a sacrifice I am willing to make".
"He gives legal work as an example, but LLM has already failed in legal work."
Only up to a point. If you read what Heller says, it can dramatically speed up discovery work or finding precedents. The link you post to failure was because the lawyers were careless enough not to check that the material which they relied upon was real. Since in complex legal cases huge amounts of time can be spent trawling for relevant material, it's still a big advance if you can set a bot to find stuff, and you then merely need to validate that content which you rely on. Moreover, using a generic ChatGPT instance to search for legal material would seem to be asking for problems, hence GoCounsel you'd expect to be better setup for the needs of legal professionals.
This entire premise is a barely veiled disclosure that they'll be charging an expert's salary for subscription access to these resources. That's why they want you to think of it as a "worker", not a SaaS expenditure.
Imagine charging 40k a year for Creative Cloud. Ludicrous, right? What about 40k per year for access to an "AI graphic designer" instead? Hey I know the visual design department is overworked and behind schedule, meet your new co-worker that we got for a steal...
Did you just land on planet earth this morning? When have you ever seen a reduction in effort within a service ever lead to cheaper prices for the service or product? If anything they push it harder to make even greater profit 'cos it costs less to get it.
The mill owners of the industrial revolution scaled this up like never before, they used machines to automated tons of processes that took skilled workers years to learn, then simply replaced them with people who just tended to the automated machines, the product could be produced faster and cheaper, so per product was cheaper but due to bulk creation of the product the profits went up by orders of magnitude. Once they'd worked this out they just pushed harder and harder.
While it's great they are checking for this, do they also look for things the AI model missed? Things that are fraud, but have been crafted more carefully, deliberately hard for an expert human lawyer to uncover.
No good if they're relying on the AI output because they're sure it doesn't hallucinate, but then miss real instances.
This is exactly how I use LLMs where I work - as a coworker. More specifically, something I can delegate tasks to, such as writing summaries of long documents, coming up with key points from brainstorm notes or meeting minutes. Luckily, we have our own garden-walled LLMs so I can feed in corporately sensitive stuff safe in the knowledge that it won't ever get out to where it shouldn't.
And before anyone asks, I never use the LLM output without reviewing it thoroughly myself. I find it most useful for rewriting existing stuff that I feed it, not for coming up with new stuff in the first place, so the "LLM Hallucination" issue is a small one and easy to deal with.
> so the "LLM Hallucination" issue is a small one and easy to deal with
Probably true - initially - as long as the fully informed resource (obviously you) is using an LLM as a tool to save time while spotting all factual errors, misunderstandings, misleading statements, missed irony and plausible, but outright made up stuff the LLM is coming up with.
Then your boss will notice that a worker with less skill/experience will produce nearly the same output given the same tools.
So after you have been replaced by 2 or 3 waves of ever cheaper resources, we arrive at a level where the untrained human basically feeds the AI and forwards its decision/analysis, if humans are still in the loop at all.
And somewhere along that process LLM hallucinations will have become a problem - but no longer for you.
Says a guy who's made millions from selling AI software.
Yes, we know you're doing that by promoting its time and labour saving capabilities, but that's been true of all software sold to businesses I can think of, so spare us the "think of it as a person" bullshit thank you - it's just software.
Yes, we recently had MS in to demo co pilot (I wasn't impressed but that's an aside), all the way through, the MS robot referred to clippy 2 as "he", it got on my nerves the first 100 times but after that I just switched off completely from the presentation.
It's clear they're trying to humanise this thing so people buy into it more but please, stop. Just stop.
Use IT and AI or lose IT and AI to IT and AI ..... and/but do take extraordinarily exceptional care to neither abuse nor misuse nor mistreat IT and AI, for then is the resultant consequence an already universally well known, known known and surprisingly effective existential threat[s] which you are once again, and surely very clearly here too on El Reg publishing situations, advised to not invoke/summon/invite/entertain for they are quite rightly to be feared and gravely to be regarded.
Do not Risk the Wrath of an IT and AI you have Zero Command and Negative Control Function over, for such tempts the Kiss of Certain Assisted Death out of the Deep Shade and Dark Shadows and into the Ever Emerging Virgin Fields of Future Work, REST and Greater IntelAIgent Games Play.
And Yes, such is/are, should it be asked, much more than just Next Generational Weapons of Mass Destruction for Global Deployment and Exercise Entertaining Human Extinction in an AI Cull of Arrogant Ignorant Dissent.
Do you think a United States Cyber Force is a Defense Imperative and good plan for activation/self actualisation?
amanfromMars [2406021747] ..... shares on https://www.nationaldefensemagazine.org/articles/2024/5/29/algorithmic-warfare-report-urges-pentagon-to-stand-up-independent-cyber-forceCyber warfare and virtual defence of national and international and internetional assets are not problems for addressing and successfully resolving with solutions involving military thinking entertaining traditional conventional punitive destructive activity.
To imagine and continue to contend that it does, guarantees one fails catastrophically against all sources and forces engaging within the fundamentally different, novel ways of ensuring and assuring friends and allies, and oneself also of course, an always overwhelming advantage.
That most certainly is one small step for a man, and probably already made some long time ago by any number of other human beings, but still possibly one giant quantum communication leap for Mankind.
What do you need to get going with the Program and ITs AI ProgramMING?
And there are not many ready, willing and/or able to tempt systems administrations with that core future nugget for processing and refinement.
* ....... Read 'Em and Weep
As an career advancement for people whose jobs were being outsourced?
The difference is, there were still jobs that weren't outsourced those people could get. Once the AI replacements are ready to take over job X, there will never be another opening for a human for job X ever again. And if AI keeps improving, the rate at which it replaces jobs will accelerate and over time far outstrip the ability for humans to learn new jobs (even if there were enough new jobs created for them)
"Step this way, yes there's a job in here for you. What is it? Oh you'll be providing something very, very valuable for the elite overlords. No ignore the sound or whirring knives behind that door, just step into this booth. Trust me within the hour you'll served up, sorry slip of the tongue! You'll be served very well indeed by taking this opportunity."
"Hey I run a company and I've not had my say on crowbarring AI into something that doesn't really need it!".
2009 - Big Data/Analytics
2014 - Cloud
2019 - Metaverse
2024 - AI
2029 - Me retiring from the tech world madhouse and volunteering more often at a local nature trust!
more lawyers?
Two AU legal academics have recently published a critique of the use of expert evidence in AU courts. They appear not to be too impressed by all the legal chaps involved reserving a particular lack of praise for the defending counsels' apparent inability to effectively cross examine expert witnesses or to critically examine their testimony.
I cannot imagine a coco unsel service added to mix could ever improve this.
Its been a fair while since most companies have actually been selling software (licenses.) Its invariably a subscription for the software or service. Cannot see anything new here.
I use LLMs for software coding assist and for processing and transforming chunks of text. They don't invalidate my job but they do make me more efficient. A 5% efficiency boost at a scale of 100.000 people means approximately 5000 fewer employees needed in order to achieve the same result. This isn't any different than a good IDE, faster compilation times, automated bug catching, and better tooling in general. The only difference is that it is happening across all fields at the same time, and that it has bigger impact on creative professions (in aggregate).
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
