[HN Gopher] Throw more AI at your problems
       ___________________________________________________________________
        
       Throw more AI at your problems
        
       Author : vsreekanti
       Score  : 59 points
       Date   : 2024-10-24 16:51 UTC (6 hours ago)
        
 (HTM) web link (frontierai.substack.com)
 (TXT) w3m dump (frontierai.substack.com)
        
       | headcanon wrote:
       | I'll stay out of the inevitable "You're just adding a band aid!
       | What are you really trying to do?" discussion since I kind of see
       | the author's point and I'm generally excited about applying LLMs
       | and ML at more tasks. One thing I've been thinking about is if an
       | agent (or collection of agents) can solve a problem initially in
       | a non-scalable way through raw inference, but then develop code
       | to make parts of the solution cheaper to run.
       | 
       | For example, I want to scrape a collection of sites. The agent
       | would at first apply the whole HTML to the context to extract the
       | data (expensive but it works), but then there is another agent
       | that sees this pipeline and says "hey we can write a parser for
       | this site so each scrape is cheaper", and iteratively replaces
       | that segment in a way that does not disrupt the overall task.
        
         | malfist wrote:
         | What do you mean the patient is bleeding out? We just need to
         | use more bandaids!
        
         | cyrillite wrote:
         | Well, the standard advice for getting off the ground with most
         | endeavours is "Do things that don't scale". Obviously scaling
         | is nice, but sometimes it's cheaper and faster to brute force
         | it and worry about the rest later.
         | 
         | The unscalable thing is often like "buy it cheap, buy it twice"
         | but it's also often like "buy it cheap, only fix it if you use
         | it enough that it becomes unsuitable". Makers endorse both
         | attitudes. Knowing when which applies is the challenging bit
        
         | westoncb wrote:
         | Cool idea. It's a bit like what happens in human brains when we
         | develop expertise at something too: start with general purpose
         | behaviors/thinking applied to new specialized task--but if that
         | new specialized task is repeated/important enough you end up
         | developing "specialized circuitry" for it: you can perform the
         | task more efficiently, often without requiring conscious
         | thought.
        
       | crooked-v wrote:
       | With the current state of "AI", this strikes me as a "I had a
       | problem and used AI, now I have two problems" kind of situation
       | in most cases.
        
         | ToucanLoucan wrote:
         | All the snark contained within aside, I'm reminded of that
         | ranting blog post from the person sick of AI that made the
         | rounds a little ways back, which had one huge, cogent point
         | within: that the same companies that can barely manage to ship
         | and maintain their current software are not magically going to
         | overcome that organizational problem set by virtue of using
         | LLMs. Once they add that in, then they're just going to have
         | late-released, poorly made software that happens to have an LLM
         | in it somewhere.
        
           | WesleyLivesay wrote:
           | I believe you mean this one:
           | https://ludic.mataroa.blog/blog/i-will-fucking-piledrive-
           | you...
        
             | Mistletoe wrote:
             | Some things should be bronzed so we don't ever lose them
             | and this is one of them.
        
           | throwaway19972 wrote:
           | > Once they add that in, then they're just going to have
           | late-released, poorly made software that happens to have an
           | LLM in it somewhere.
           | 
           | Hell, you can even predict where by looking at the lowest-
           | paid people in the organization. Work that isn't valued today
           | won't start being valued tomorrow.
        
         | zonethundery wrote:
         | It is too bad Erik Naggum did not live to see the AI era.
        
         | paxys wrote:
         | But see the company selling the AI "solution" made money, which
         | is the entire point.
        
           | satisfice wrote:
           | Making money is the Silicon Valley Bro equivalent of
           | mathematical proof.
           | 
           | Sorry, not making money-- I meant attracting investment.
           | 
           | AI tech is generally undertested and experimental, yet
           | marketed as solid and reliable.
        
             | rachofsunshine wrote:
             | Speaking as someone who is very critical of the general
             | "raise infinite money and grow grow grow" approach to
             | startups (it's a big part of why my company is bootstrapped
             | [1]), I think there's a really important difference between
             | _making_ money and _raising_ money.
             | 
             |  _Raising_ money is about convincing investors that you can
             | solve someone else 's problem. Or even more abstractly,
             | about convincing investors that you can convince other
             | investors that you can convince other investors that you
             | can solve someone else's problem. That's enough layers of
             | indirection that signaling games start to overtake concrete
             | value a lot of the time, and that's what gets you your
             | Theranoses and your FTXes and the like. You get into "the
             | market can remain irrational longer than you can remain
             | solvent", or its corollary, "the market can remain
             | irrational long enough for you to exit before it wakes up".
             | 
             | But _making_ money means you are solving a problem _for
             | your customer_ , or at least, that your customer thinks
             | you're solving a problem with no additional layers of
             | indirection. And if we take "making money" to mean "making
             | a profit", it also means you're solving their problem at
             | less cost than the amount they're willing to pay to solve
             | it. You are actually creating net value, at least within a
             | sphere limited to you and your customer (externalities, of
             | course, are a whole other thing, but those are just as
             | operative in non-profitable companies).
             | 
             | I think this is one of the worst things about the way
             | business has done today. Doing business, sustainably and
             | profitably, is an excellent way to keep yourself honest and
             | force your theories to actually hold up in a competitive
             | market. It's a good thing for you and for your users. But
             | business has become so much about gathering sufficient
             | capital to do wildly anticompetitive things and/or buy
             | yourself preferential treatment that we're losing that
             | regulating force of honesty.
             | 
             | [1] see my HN profile for more on that if you care
        
         | o11c wrote:
         | Incidentally, this just led me to the first case I've found of
         | a wide variety of AIs actually generating the same answer
         | consistently. They almost always said a close synonym for "now
         | I have a solution".
         | 
         | For some AIs, if you ask them to complete it for "Java" and/or
         | "Regexes" first, then they give realistic answers for "AI". But
         | others (mostly, online commercial ones) are just relentlessly
         | positive even then.
         | 
         | Prompting to complete it for "Python" usually remains positive
         | though.
        
       | Stoids wrote:
       | We aren't good at creating software systems from reliable and
       | knowable components. A bit skeptical that the future of software
       | is making a Rube Goldberg machine of black box inter-LLM
       | communication.
        
         | TZubiri wrote:
         | I'm pretty sure this is a satire post
        
           | pizza wrote:
           | Here's a practical in this vein but much simpler - if you're
           | trying to answer a question with an LLM, and have it answer
           | in json format within the same prompt, for many models the
           | accuracy is worse than just having it answer in plaintext.
           | The reason is that you're now having to place a bet that the
           | distribution of json strings it's seen before meshes nicely
           | with the distribution of answers to that question.
           | 
           | So one remedy is to have it just answer in plaintext, and
           | then use a second, more specialized model that's specifically
           | trained to turn plaintext into json. Whether this chain of
           | models works better than just having one model all depends on
           | the distribution match penalties accrued along the chain in
           | between.
        
             | from-nibly wrote:
             | So developing solutions with ai is like trying to build
             | stuff with family feud.
        
         | paulddraper wrote:
         | Not only is it the future of software, it's the past and
         | present as well.
        
         | dsv3099i wrote:
         | As someone who makes HW for a living, please do make more Rube
         | Goldberg machines of black box LLMs. At least for a few more
         | years until my kids are out of college. :)
        
         | mmcdermott wrote:
         | I could see software having a future as a Rube Goldberg machine
         | of black box AIs, if hardware is cheap enough and the AIs are
         | good enough. There was a scifi novel (maybe "A Fire Upon the
         | Deep"?) where there was no need to write software because AI
         | could cobble any needed solution together by using existing
         | software and gluing it together. Throwing cycles at deepening
         | layers was also something that Paul Graham talked about in the
         | hundred year language (https://paulgraham.com/hundred.html).
         | 
         | Now, whether hardware is cheap enough or AI is smart enough is
         | an entirely different question...
        
       | l5870uoo9y wrote:
       | RAG doesn't necessarily give the best results. Essentially it is
       | a technically elegant way to semantic context to the prompt (for
       | many use cases it is over-engineered). I used to offer RAG SQL
       | query generations on SQLAI.ai and while I might introduce it
       | again, for most use cases it was overkill and even made working
       | with the SQL generator unpredictable.
       | 
       | Instead I implemented low tech "RAG" or "data source rules". It's
       | a list of general rules you can attach to a particular data
       | source (ie database). Rules are included in the generations and
       | work great. Examples are "Wrap tables and columns in quotes" or
       | "Limit results to 100". It's simple and effective - I can execute
       | the generate SQL again my DB for insights.
        
         | trhway wrote:
         | Reminds how few years ago Tesla (if i remember - Karpaty)
         | described that in Autopilot they started to extract the 3rd
         | model and use it to explicitly apply some static rules.
        
         | simonw wrote:
         | What do you mean by "RAG SQL query generations"? Were you
         | searching for example queries similar to the questions the
         | user's asked and injecting those examples into the prompt?
        
         | com2kid wrote:
         | RAG without preprocessing is almost useless, because unless you
         | are careful, RAG gives you all the weaknesses of vector DBs,
         | with all the weaknesses of LLMs!
         | 
         | The easiest example I can come up is imagine you just dump a
         | restaurant menu into a vector DB. The menu is from a hipster
         | restaurant and instead of having "open hours" or "business
         | hours" the menu says "Serving deliciousness between 10:00am and
         | 5:00pm"
         | 
         | Naive RAG queries are going to fail miserably on that menu.
         | "When is the restaurant open?" "What are the business hours?"
         | 
         | Longer context lengths are actually the solution for this
         | problem, when context is small enough and the potential of
         | ambiguity is high enough, LLMs are the better tool.
        
         | fsndz wrote:
         | Semantic search is a powerful tool that can greatly improve the
         | relevance and quality of search results by understanding the
         | intent and contextual meaning of search terms. However, it's
         | not without its limitations. One of the key challenges with
         | semantic search is the assumption that the answer to a query is
         | semantically similar to the query itself. This is not always
         | the case, and it can lead to less than optimal results in
         | certain situations. https://fsndzomga.medium.com/the-problem-
         | with-semantic-searc...
        
           | mvdtnz wrote:
           | Don't post AI slop on HN comments please.
        
       | keeganpoppen wrote:
       | YES (although i'm hesitant to even say anything because on some
       | level this is tightly-guarded personal proprietary knowledge from
       | the trenches that i hold quite dear). why aren't you spinning off
       | like 100 prompts from one input? it works great in a LOT of
       | situations. better than you think it does/would, no matter your
       | estimation of its efficacy.
        
         | pizza wrote:
         | 100 prompts doing what? Something like more selective, focused
         | extraction of structured fields?
        
       | hggigg wrote:
       | I wish this was funny but it's not. We are doing this now. It has
       | become like "because it's got electrolytes" in our org.
        
         | glial wrote:
         | Well, at least it's not blockchain or Kubernetes.
        
           | pqdbr wrote:
           | The blockchain hype train was ridiculous. Textbook "solution
           | looking for a problem" that every consultant was trying to
           | push to every org, which had to jump onboard simply because
           | of FOMO.
        
             | fluoridation wrote:
             | I don't think that's quite right. It was businesses who
             | were jumping at consultants to see how they stuff a
             | blockchain into their pipeline to do the same thing they
             | were already doing, all so they could put "now with
             | blockchain!" on the website.
        
           | hggigg wrote:
           | We did those already. So still not funny :)
        
       | mvdtnz wrote:
       | We are truly in the stupidest phase of software engineering yet.
        
       | com2kid wrote:
       | > This is where compound systems are a valuable framework because
       | you can break down the problem into bite-sized chunks that
       | smaller LLMs can solve.
       | 
       | Just a reminder that smaller fine tuned models are just as good
       | at solving the problems they are trained to solve, as large
       | models are.
       | 
       | > Oftentimes, a call to Llama-3 8B might be enough if you need to
       | a simple classification step or to analyze a small piece of text.
       | 
       | Even 3B param models are powerful now days, especially if you are
       | willing to put the time into prompt engineering. My current side
       | project is working on simulating a small fantasy town using a
       | tiny locally hosted model.
       | 
       | > When you have a pipeline of LLM calls, you can enforce much
       | stricter limits on the outputs of each stage
       | 
       | Having an LLM output a number from 1 to 10, or "error" makes your
       | schema really hard to break.
       | 
       | All you need to do is parse the output and it if isn't a number
       | from 1 to 10... just assume it is garbage.
       | 
       | A system built up like this is much more resilient, and also
       | honestly more pleasant to deal with.
        
       | fsndz wrote:
       | More AI sauce won't hurt right ? Meanwhile, we still have to
       | solve the ragallucination problem. https://www.lycee.ai/blog/rag-
       | ragallucinations-and-how-to-fi...
        
       | kylehotchkiss wrote:
       | Use Moore's law to achieve unreal battery life and better
       | experiences for users... or use Moore's law to throw more piles
       | of abstractions on abstractions where we end up with solutions
       | like Electron or I Duck Taped An AI on it.
       | 
       | Reading through this, I could not tell if this was a parody or
       | real. That robot image slopped in the middle certainly didn't
       | help.
        
       ___________________________________________________________________
       (page generated 2024-10-24 23:01 UTC)