Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI and the Automation of Work

ben-evans.com

Post reply on HN
Print view View original

21–30 of 268 posts

Re: AI and the Automation of Work

#21

Post by famouswaffles »

skepticATX wrote: post #18
famouswaffles wrote: post #15

Earlier quoted context omitted.

Current Capability would be the biggest one. We're at the point where any testable definitions of GI that the sota LLM fails (GPT-4) is also failed by a good chunk of humans. You couldn't say that a few years ago nevermind 60. What we have now (so no hypotheticals) coupled with the fact that scaling hasn't yet shown any performance walls makes a pretty good shout that things will probably be different this time.

That's not really true though. LLMs are abysmal at planning, for example. Something that comes quite naturally to humans.

They're really only abysmal if you attempt to one-shot it and probe with tasks that would require a human a scratchpad to accomplish.

Humans can't one-shot non trivial planning tasks either. It's the one problem i have with all the papers that try to evaluate planning for LLMs.

Step away from that approach and they're ok.

https://innermonologue.github.io/

https://tidybot.cs.princeton.edu/

Re: AI and the Automation of Work

#22

Post by danielmarkbruce »

runsWphotons wrote: post #19
skepticATX wrote: post #18

Earlier quoted context omitted.

That's not really true though. LLMs are abysmal at planning, for example. Something that comes quite naturally to humans.

They are probably better than 10% of people

You meant 70%, right?

Re: AI and the Automation of Work

#23

Post by 29athrowaway »

mdale wrote: post #6

I think the "sparks of AIG" presentations are compelling. It does feel like there are a just a few remaining big problems away from AGI level capabilities where it was more futuristic unclear projections in the past. If the issues as synthetic data conversations like alpha go but for LLMs; setting much longer token contexts (or dramatically cheaper training on custom data sets) And scaling via software and hardware a…

Yann LeCun says human intelligence is not "general intelligence".

Re: AI and the Automation of Work

#24

Post by pulvinar »

skepticATX wrote: post #18
famouswaffles wrote: post #15

Earlier quoted context omitted.

Current Capability would be the biggest one. We're at the point where any testable definitions of GI that the sota LLM fails (GPT-4) is also failed by a good chunk of humans. You couldn't say that a few years ago nevermind 60. What we have now (so no hypotheticals) coupled with the fact that scaling hasn't yet shown any performance walls makes a pretty good shout that things will probably be different this time.

That's not really true though. LLMs are abysmal at planning, for example. Something that comes quite naturally to humans.

I'm curious as to your source, or particular examples, since they (or at least GPT-4) seem to me to be rather decent at planning. E.g., for writing code.

Re: AI and the Automation of Work

#25

Post by fbdab103 »

chologrande wrote: post #3

What exactly is a "shadow IT app" ? How in the heck is _Evernote_ the top shadow IT app or related to automation and aI?

That entire list is bananas. Not one mention of PowerAutomate or BI? Sticking within the typical Microsoft stranglehold offers a fair amount of tooling with multiple integration points. They come bundled with everything else in a corporate environment, so they are "free" and less intimidating for a non-technical person to dip their toe into than any other low-code tool.

Re: AI and the Automation of Work

#26

Post by DougEiffel »

add-sub-mul-div wrote: post #10
somewhereoutth wrote: post #9

Or it could be that the time/effort required to craft a prompt and check the output is not worth it for most/all use cases, and thus after the initial excitement ChatGPT and friends gather dust much like blockchain and Apple Vision.

Yeah this hype has opened my eyes to the fact that there are so many people who haven't already found a flow for internet searches that's more streamlined than this AI stuff. Don't we all do this every day?

I'll be the odd man out here - I teach special education and GPT has been a lifesaver. I feed it a complex task or text, ask it to simplify it down to my students' reading levels, and then just do a super quick read/edit.

Similarly, I've been using Dall-E to add pictures to the books we read in class.

Compare this to retyping everything at a 4th grade reading level and then painstakingly searching Google for pictures (and usually ending up with mismatched pictures with varying artistic styles or watermarks).

I don't know that we will have true artificial intelligence soon a lá I Robot, but I feel nervous and optimistic about what's to come either way.

By the way, call me crazy but I think the Apple Vision thing is going to be a game changer for business and consumers.

Re: AI and the Automation of Work

#27

Post by blueblimp »

acover wrote: post #2

> As an analyst, though, I tend to prefer Hume’s empiricism over Decartes - I can only analyse what we can know. We don’t have AGI, and without that, we have only another wave of automation, and we don’t seem to have any a priori reason why this must be more or less painful than all the others.

This is a good way to look at it. The deployment of the LLM technology we currently have is one thing. The impact of AGI, when it's created in the future, is a different thing. And people are often not clear on which of the two they're discussing.

Re: AI and the Automation of Work

#28

Post by benedictevans »

fbdab103 wrote: post #25
chologrande wrote: post #3

What exactly is a "shadow IT app" ? How in the heck is _Evernote_ the top shadow IT app or related to automation and aI?

That entire list is bananas. Not one mention of PowerAutomate or BI? Sticking within the typical Microsoft stranglehold offers a fair amount of tooling with multiple integration points. They come bundled with everything else in a corporate environment, so they are "free" and less intimidating for a non-technical person to dip their toe into than any other low-code tool.

This list is shadow IT, which means the things that the corporate IT department isn't paying for and often doesn't even know about. The tools you mention are in the Productiv (and Okta) data in a different category.

Re: AI and the Automation of Work

#29

Post by Given_47 »

Aperocky wrote: post #20
kelvinjps wrote: post #4

I wonder if a unix like worlflow will be popular again, in general, I mean is easier to automate, ask it to make scripts etc

Probably won't, people don't ask for what they don't know, and you'd be surprised at how even in the field of software engineering people don't actively looks for ways to automate stuff.

Then what use cases are the general public using LLMs for other than automation for it to have exploded in popularity? Tho anecdotally it seems to be used for “wisdom” of the crowds coupled with blind trust in the objectivity of a language model

Re: AI and the Automation of Work

#30

Post by climatologist »

pulvinar wrote: post #24
skepticATX wrote: post #18

Earlier quoted context omitted.

That's not really true though. LLMs are abysmal at planning, for example. Something that comes quite naturally to humans.

I'm curious as to your source, or particular examples, since they (or at least GPT-4) seem to me to be rather decent at planning. E.g., for writing code.

Ask it to write a backtracking sudoku solver with coroutines and/or fibers and let me know how it performs in your language of choice.

We are nowhere near generally intelligent software systems.

Post reply on HN