I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall Alan Kay has (re)framed things this way as well. I believe he described the important skill as "Problem recognition"
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall This really matches my experience interviewing for jobs lately: the coding tests are often very artificial problems that you’ll never encounter in real work situations and that you could answer quickly, under the intense pressure of an interview, only if you had seen that exact problem before or something very similar and learned the solution by heart. The interviewers always say, ‘We just want to see how you think’, but in practice they’re testing how much you’ve memorised.
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall I think though, it makes sense for them: in a large enough internal repository generated by thousands of engineers, you'll likely find a problem/solution near enough to yours, and reusing/adapting instead further proliferating has theoretic maintainance value.
Not saying this is what actually happens (often it is genuinely easier to roll your own than adapt/share something existing).
Then there are always the problems or people you can't contain, generating enough novelty.
-
@david_chisnall This really matches my experience interviewing for jobs lately: the coding tests are often very artificial problems that you’ll never encounter in real work situations and that you could answer quickly, under the intense pressure of an interview, only if you had seen that exact problem before or something very similar and learned the solution by heart. The interviewers always say, ‘We just want to see how you think’, but in practice they’re testing how much you’ve memorised.
@benjamingeer @david_chisnall and how strong the PTSD is from those memories
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall oh well the Adeptus Mechanics will never learn
️ -
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
Again very good enlightening words. Except that for this, I already knew it before. Maybe I have grown some wisdom?
The thing is, those companies are founded to make profit. To make profit you can go the hard way and create a value for the society, or go the easy way and create problems in the minds of people then offer your solutions.
What I'm trying to point out is that many needs people have today aren't "real" needs. Corpos creates solutions and convinced people they have that need and the solution is at their hands.
Also, sometimes the companies themselves create the problems. This reminds me of a movie animation:
"When we start manufacturing these fresh air bottles, the air quality will worsen. Then people will come to us to buy the bottles to have fresh air"
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall Silibandia built an entire culture around pattern matching. Instead of genuine human ships. True optimocracy requires identifying the right problems to solve, not just running automated, pattern-matched scripts on existing ones.
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall This is such an interesting description, I need to mull it over for a few days to fully process it
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
I've worked out that a lot of IT types think that if they model a process, they can automate it, for a more "efficient" result. Because they are super logical and very clever.
They never question whether the process needs to exist, nor why it's done so inefficiently when it does. Often the correct answer is "regulations and legislation".
Now you know why things are going the way they are.
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall I think you might be underestimating the value here. Just with "matching solved problems with existing solutions" in it's own right is the greatest invention of the century. Most problems have existing solutions - choosing the right one and applying it is like 99% of all work we do. This is actually a good thing as it leaves creative edge pushing solutions for us and have LLM tools do the "boring" matching - is it not?
-
@david_chisnall I think though, it makes sense for them: in a large enough internal repository generated by thousands of engineers, you'll likely find a problem/solution near enough to yours, and reusing/adapting instead further proliferating has theoretic maintainance value.
Not saying this is what actually happens (often it is genuinely easier to roll your own than adapt/share something existing).
Then there are always the problems or people you can't contain, generating enough novelty.
@robinp @david_chisnall And yet, only once in an interview was I commended for being able to *find* the answer (actual skill being tested) instead of for knowing the answer.
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall Good post. Seems like the intro to a lot more thoughts on the subject!
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall I see it as simple end stage capitalism
If you sell people tools they compete with you. The tech industry aspired to sell you closed boxes that did only what they wanted, and even if third parties were involved only by paying a rental in tools the masses were not permitted.
The gaming world got away with it for a time and other markets have on and off.It's not new, it's not even silicon valley - it's the same mindset as selling food and stopping seed access.
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
I think LLMs demonstrate a huge amount of resource consumption and - with the market being saturated with compute - big tech tries to push it onto everybody, so companies buy new hardware to participate in the inflated hype with big $$$.
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall so, applied mathematicians are what they're looking for. It would explain why I see so many "physicists who enjoy coding" in the field. I would have changed my entire focus had I known that in uni. But my former university thought it this "great thing" if you, as a student of maths, became an actuary. 🥱
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall From your "infer from nearby" comment it seems you believe LLMs work differently to how they actually do.
Maybe it's not that obvious in software development (since there exists no development where it's not easy to see the shoulders being stood upon) but some of the problem solving that has been done in Maths should.
-
Again very good enlightening words. Except that for this, I already knew it before. Maybe I have grown some wisdom?
The thing is, those companies are founded to make profit. To make profit you can go the hard way and create a value for the society, or go the easy way and create problems in the minds of people then offer your solutions.
What I'm trying to point out is that many needs people have today aren't "real" needs. Corpos creates solutions and convinced people they have that need and the solution is at their hands.
Also, sometimes the companies themselves create the problems. This reminds me of a movie animation:
"When we start manufacturing these fresh air bottles, the air quality will worsen. Then people will come to us to buy the bottles to have fresh air"
@farooqkz @david_chisnall you just described the "CyberSecurity" industry. Congratulations!
-
@david_chisnall This really matches my experience interviewing for jobs lately: the coding tests are often very artificial problems that you’ll never encounter in real work situations and that you could answer quickly, under the intense pressure of an interview, only if you had seen that exact problem before or something very similar and learned the solution by heart. The interviewers always say, ‘We just want to see how you think’, but in practice they’re testing how much you’ve memorised.
@benjamingeer yeah because the recruiters are incompetent or don't care enough to engage.
Imagine if you were hired by your future colleagues? A team interview would ask you practically useful questions and you would have equal opportunity to see you really wanted to work with these people.
No managers, no CEOs, no mental constipation.
But this won't happen because we are hired to make money for some asshat who couldn't give less of a shit about you or your colleagues, making something that nobody gives a shit about.
And so nobody gives a shit, and the few folks who come with passion slowly (a) stop giving a shit, (b) burn out, or (c) turn traitor to become asshats themselves.
Capitalism! Giving you the middle finger. Suck it.
Don't like it? Do something! But if you do anything that gives people hope, better buy some guns and get ready to kill a fuckton of cops, and then soldiers. (This has happened in Europe many times. Either the cops or the military wins.)
Meanwhile, do as little as possible and safely sabotage whatever you can. Or join a worker's coop! Where are those??
-
@david_chisnall I think you might be underestimating the value here. Just with "matching solved problems with existing solutions" in it's own right is the greatest invention of the century. Most problems have existing solutions - choosing the right one and applying it is like 99% of all work we do. This is actually a good thing as it leaves creative edge pushing solutions for us and have LLM tools do the "boring" matching - is it not?
It's how I use LLMs. Old greybeard, can "code anything" - but that doesn't mean that I _want_ to. When I have an itch to scratch I constantly have to weigh whether I have the time and interest to do _all_ of the things around the itch or not.
With LLMs I no longer have to ponder. I do the tiny fun part and the LLM does the rest - under my very greybeardy guidance.
I have never before created so many itch-scratching open source projects as during the last few months.
-
I see LLMs as a continuation of the Silicon Valley obsession with 'problem solving' and their specific definition.
Back when I was an undergrad, the refrain we heard from big tech recruiters was 'we're not interested in specific skills, we want to hire problems solvers'. This always struck me as odd: everyone in my year could solve problems, it wasn't a rare skill, the difficult thing was identifying the correct problems to solve.
Digging a bit deeper, it turns out that they didn't even want people capable for solving generic problems, they wanted people who had learned a load of problem-solution pairs and would do closest-fit matching. People who would look at a problem and say 'ah, this looks like this well-known problem, the solution is therefore a variation of this well-known solution'. Not people who would create novel solutions but people who could pattern match and apply off-the-shelf solutions with small tweaks.
They wrote books about how to hire people with that skill, all without being explicit that this is what they were looking for.
And now they have machines that can do this: ingest a load of problem-solution pairs and match problems to some space of problems and infer a solution from the nearby solutions.
It's not surprising that they believe this is the same as novelty, because they've spent three decades incentivising their employees to avoid true novelty.
@david_chisnall funny how experiences change. A passing insult that stuck with me was someone disparaging older tech workers who describe themselves as "problem solvers" (which I did). The recruiter said this was an indicator that they have no actual skills to list.
