If the problem as stated is "Performing an LLM query at newly inflated cost $X is an iffy value proposition because I'm not sure if it will give me a correct answer" then I don't see how "use a tool that keeps generating queries until it gets it right" (which seems like it is basically what you are advocating for) is the solution.
I mean, yeah, the result will be more correct answers than if you just made one-off queries to the LLM, but the costs spiral out of control even faster because the agent is going to be generating more costly queries to reach that answer.
Which isn't the same as saying LLMs and related technology aren't useful... they are.
But as you mentioned the financials don't make sense today, and even worse than that, I'm not sure how they could get the financials to make sense because no player in the space on the software side has a real moat to speak of, and I don't believe its possible to make one.
People have preferences over which LLM does better at job $XYZ, but I don't think the differences would stand up to large price changes. LLM A might feel like its a bit better of a coding model than LLM B, but if LLM A suddenly cost 2x-3x, most people are going to jump to LLM B.
If they manage to price fix and all jump in price, I think the amount of people using them would drop off a cliff.
And I see the ultimate end result years from now (when the corporate LLM providers might, in a normal market, finally start benefiting from a cross section of economies of scale and their own optimizations) being that most people will be able to get by using local models for "free" (sans some relatively small buy-in cost, and whatever electricity they use).
Its more about abolishing the premise of the Senate where every state gets 2 senators regardless of their population.
(Though the filibuster issue is also a valid debate lately)
The founders had decent intentions for this design, but I'm fairly sure the vast majority of them would have changed their mind if they knew just how concentrated the population of the US would end up and how the system would act to give the minority far too much power rather than protect them from having too little.
Ah ok, I hadn't noticed there was also recent discussion on the Senate itself, wrt not being numerically representative.
People often say stuff like "the founders would have changed their mind if they knew just how concentrated the population would end up [wrt representation]", but they don't propose anything specific or constructive (short of federal-state litigation, secession or another civil war). How about a (neutral) commission to reapportion State boundaries every 10 years based on Census results (with some population formula between not-quite-linear and wildly disproportionate)? Or else, to periodically reapportion state counts of Senators to total 100. (Obviously these couldn't get ratified these days, but they just might have in the 1790s). If not, what's your specific suggestion?
Another thing people aren't currently discussing much is how badly break down if/when the Supreme Court gets captured by a dominant group that is both ideological and not independent. Look at how high the stakes will be for nominating the eventual replacement to Justice Clarence Thomas/Sotomayor/etc.
And of course the terrible Citizens Utd ruling muddies every consideration of representation.
And then there's also the parallel discussion of the Senate filbuster rule, remember though that if there was no filibuster, Citizens Utd would allow unlimited dark money to influence every vote, specifically all the action would focus on the Senators in the middle, think Joe Lieberman, Olympia Snowe, Susan Collins, Manchin, Sinema. Seems near-impossible to reconstruct democracy under these constraints. (Look at the recent Senate stealth attack in the shutdown bill by lobbyists for newly-legalized CBD to try to ban Hemp).
Do you believe there's a distinction between giving the minority too much power versus protecting them from having too little? It seems like the same thing said two different ways.
Giving the minority (Christian extremists in flyover states) too much power has caused them to start revoking the rights of everyone not them. Forget about protecting minority groups from having too little power, the main concern now is wresting control of our government in line with the principle of "one person, one vote" instead of "hectares of corn and churches take precedence over people".
> OK then I guess we should disconnect every large industrial electricity customer from the grid.
No, what we should do is put every "large load" electricity customer (including but not limited to these data centers) into their own rate-payer class like they did in Maryland and Oregon instead of lumping them in with everyone else.
Once they are in their own rate-payer class then their rates can be adjusted to pay for the costs of the increased infrastructure that is only needed because they exist (take away data center build-outs and electricity usage is largely flat or falling pretty much everywhere in the US).
I hope the data center developers are paying you to lobby for their ongoing corporate welfare? Because that's what you're basically doing here.
If they pay the same rates as everyone else then that's not corporate welfare. If the rates are artificially low thus causing shortages then we have a different problem.
> If they pay the same rates as everyone else then that's not corporate welfare.
It is, because the electricity companies don't have magic electricity generating machines that can scale infinitely.
To satisfy the new demand which only exists because the data center was built, they spend a lot of money on new infrastructure. They then raise everyone's prices by an equal percentage to support this new infrastructure even though the infrastructure was not needed until the data center was built.
Not charging the data center developers for that extra build out and expecting everyone to absorb the costs for new infrastructure that never would have been built if the data center wasn't built is absolutely corporate welfare.
MAUI was horrible when I tried using it about a year ago, tons of bugs, pretty iffy comms/support from the MAUI team as to timelines when things might get fixed, etc.
Eventually dumped it and moved to Kotlin Compose/Multiplatform, which is just so much better at achieving a similar goal (though, obviously, without being part of the .NET ecosystem).
I did nerdy kid stuff like read lots of books and use computer screens (uncommon back then, I'm 52)... but I also played outdoors for hours nearly every single day.
Ended up with vision at -6.00 by the time I was a young teenager (don't remember the age it started to slide that way but would estimate around 7-8 or so). Hasn't gotten any worse (or better) since then.
I'm sure they would be if the stock price had ever showed any signs of being based in reality.
But for now Elon can keep having SpaceX and xAI buy up all the unsold Teslas to make number go up.
If that ever stops working, just spin up a new company with a hyper-inflated valuation and have it acquire Tesla at some made up number. Worked for him once, why not try it again.
And at this point he can get even fraudier, with the worst possible realistic outcome being that he might get forced to pay a relatively small bribe and publicly humiliate himself for Trump a bit.
But there's really no more consequences to any sort of business fraud (for now) as long as you can afford the tribute.
The current tech job situation in the US is hitting junior developers a lot harder than senior developers, which is why there's a general assumption a lot of it is being driven by the combination of AI and outsourcing.
Whether this is sustainable remains to be seen, there was a big outsourcing trend back in ~2004 in the tech industry here that ended up being somewhat short-lived as many companies realized those efforts were costing them more than they were saving them beyond the short term. Whether or not this time is different with the added AI component, I have no idea. I wouldn't bet on it in either direction.
Its not great out there for senior developers either, but on the senior side its more of a freeze (try not to lose your job because the next one may be very hard to find) whereas on the junior side its more of a clear contraction (keeping your job will be much harder, finding a new one harder still).
> You have an amazing amount of confidence that the new devs are going to fix subtle race conditions…
> They’ll add a few sleep(5) calls to make them go away though..
I don't think torginus's point is that the new devs will find the proper fixes for the code, more that such a hack might be good enough in the eyes of both the company's management and the company's users.
As much as it pains me to recognize this (as a fan of clean, elegant code) not every bit of software needs to be clean and elegant to achieve its intended purpose (which is, at least in the corporate software world being discussed here, to make money).
If you meet the needs of your software's users, you can make a lot of money for a lot of years selling a piece of crap held together with chewing gum and elastic bands (and many companies have).
Such "sleep(5)" is actually a milestone in a project, a milestone that marks the beginning of deterioration and the end of architectural changes. I've seen multiple pull requests with "sleep" and similar shortcuts, and worse, I've rewound coding agent context and changes because of such model suggestions. I'm responsible for informing management about the consequences of such shortcuts and why we have to take the correct, often more time-consuming and more expensive, approach to avoid project derailment in the long term. I believe that management picks employees. If they don't trust my judgment, then it's okay for me, but I don't feel obliged to be responsible for the consequences.
I'm certainly not defending such code, just saying that its out there and in some very successful projects.
If the managers where you work understand the dangers of taking on such tech debt and are willing to put in the resources to avoid it, consider yourself lucky, because in my experience that's not at all universal, or even particularly common, though it does certainly exist at some places.
Unfortunately the Bivens act was already heavily neutered well before this current trainwreck of a Supreme Court we have now was fully assembled (saving them the trouble of having to neuter it themselves):
If the problem as stated is "Performing an LLM query at newly inflated cost $X is an iffy value proposition because I'm not sure if it will give me a correct answer" then I don't see how "use a tool that keeps generating queries until it gets it right" (which seems like it is basically what you are advocating for) is the solution.
I mean, yeah, the result will be more correct answers than if you just made one-off queries to the LLM, but the costs spiral out of control even faster because the agent is going to be generating more costly queries to reach that answer.
reply