"hand over" is a misnomer - what actually happens is that there's an interaction with a machine and people either trust it too much, or forget that it's a machine (i.e. handed from one person to another and the "AI warning" label is accidentally or intentionally ripped off)
Two things are holding back current LLM-style AI of being of value here:
* Latency. LLM responses are measured in order of 1000s of milliseconds, where this project targets 10s of milliseconds, that's off by almost two orders of magnitute.
* Determinism. LLMs are inherently non-deterministic. Even with temperature=0, slight variations of the input lead to major changes in output. You really don't want your DB to be non-deterministic, ever.
From what I understand, in practice it often is true[1]:
Matrix multiplication should be “independent” along every element in the batch — neither the other elements in the batch nor how large the batch is should affect the computation results of a specific element in the batch. However, as we can observe empirically, this isn’t true.
In other words, the primary reason nearly all LLM inference endpoints are nondeterministic is that the load (and thus batch-size) nondeterministically varies! This nondeterminism is not unique to GPUs — LLM inference endpoints served from CPUs or TPUs will also have this source of nondeterminism.
"But why aren’t LLM inference engines deterministic? One common hypothesis is that some combination of floating-point non-associativity and concurrent execution leads to nondeterminism based on which concurrent core finishes first."
great post - let me add that native was forced into some of this by the web:
1. locked up files ==> that's for security, which wasn't an issue in the 1990s.
2. inconsistent look ==> people are embedding browsers inside apps for all sorts of reasons, ruining the "native" UI/UX even if the OS "look" were stable.
It took a while, but once again open source and the web kinda won, though if you like consistency, then I agree it's a pyrrhic victory...
so that's not true - I worked for years in the grocery business and prices DO come down and in fact, I've seen evidence all over the NYC market of prices falling recently.
examples include eggs for $2.99 in some places (!), and other competitive categories like unbranded meat and cheese, pasta, and more.
prepared foods seem to be slower, I'm assuming because labor costs continue to rise.
"That's not true" is too strong a statement on your part.
The statistic you cite does not necessarily contradict what the parent comment is saying. "Up 29% since February 2020" is an absolute change since a specific point. The parent comment is saying prices have "come down" i.e. since their peak. It can still be up overall, so long as it's not up as high as it was at one point.
EDIT: To be clear, the parent comment might still be wrong, or might be right only within a biased sample (i.e. their own experience). I'm only making the point that the statistic you're referencing does not outright disprove what they're saying. Prices can be up since six years ago AND down since two years ago (random time periods chosen for illustration only).
Of course this is talking about the overall price level. The prices in specific sectors can fluctuate independently of that. Food and energy in particular are excluded from core inflation because they're especially volatile.
Prices never came back to pre-pandemic levels, that is absolutely correct. But if you remember that prices ballooned last year when Trump just took office, eggs were getting more and more expensive, etc and I gotta say prices came down a bit after that, but always never to previous levels.
https://ando.work/ is revolutionizing life for 100Ms of hourly workers with shift schedules and assignments that fit their lives, their families, second+ jobs and hobbies, personalities, commutes, etc.
We're looking for a lead data/pipeline/integrations engineer who can help connect to POS, HR, ERP, etc. While LLMs will write the code of course, we need someone who understand the gestalt of integration, clean data, working with sometimes-difficult partners, workarounds for messy situations, scaling, etc.
We're also looking for algorithms engineers with experience in scheduling and assignment, but excited to take an LLM-first approach, which we've proven works in this case and blows away classical solvers. Yes, we've done the math on scaling.
Ando is a Seed stage startup, backed by SV insiders with delightful track records.
Long ago, my next door neighbor's daughter had severe SB and was confined to a wheelchair, slow mental and emotion development, etc. Nobody thought she'd live, but in fact got to adulthood. It was basically a full-time job for her (single) mom.
Sadly, the divorce rate is over 50% for parents of spina bifida children. My daughter is almost 7 and my wife and I are happy, we have a good life and I’m fortunate to have a career that lets her stay at home to tend to our daughters multiple monthly doctors appointments, and one where I can largely work from home to help when she needs carrying (which is often!)
That said, our daughter is a miracle, she’s intellectually sharp, and we do everything we can to give her all the experiences any other bright and outgoing child would have. I hold her up so she can play the games she can’t reach at Dave and Busters, even if my arms get tired. We go on hikes with an expensive stroller that’s also a bike made for off road biking because we read Little House and she wanted to know what prairie looked like. We plan on getting an RV to take her to national parks.
Her wheelchair tennis coach recently tried an exoskeleton that allows him to walk at a research lab in New York, she was elated. She asks me when her robot legs are coming. I tell her we don’t know but robots in every house would certainly help that sort of technology move forward. I tell her “they’ve got to test it on adults before kids can get one!”
When she was one, they told us to make plans for the future, to get our affairs in order. Three years later the palliative care worker who had told us our child wouldn’t live past her second birthday came to visit us during a hospital visit and talk to us, so happy to see a case where they’d all been wrong. I’m so happy they were.
lol, nothing to do with Biden - Trump soiled his brand in NYC over 4+ decades of screwing people over, not paying bills around town, "strategic bankruptcies" etc.
It's telling that "trump" buildings rebranded in NYC...
reply