lumost

14 ms·

301.
by lumost 1y ago
Usually these mandates are made by someone who evaluates “risks.” Third party risks are evaluated under the assumption that everything will be done sensibly in the 1p scenario, to boot, the 1p option will be cheaper as disk drives etc are o
302.
by lumost 1y ago
The winner takes all thesis would be that like TSMC, the capex of competing in this field keeps growing until only one vendor can both raise sufficient capital to compete and effectively execute with that capital. OpenAI doesn't need t
303.
by lumost 1y ago
There are multiple issues in the Phd/Faculty market. Different fields are at different phases of the cycle. The biggest trend is that Universities expanded in the 1950s and 60s, and are now contracting as population declines. Universit
304.
by lumost 1y ago
This does point to one of the problems the article highlights. The US graduates 1385 PhDs in economics for 400 faculty positions. Those faculty positions face global competition. While it's a net benefit to US institutions to have the
305.
by lumost 1y ago
OpenAI has a reasonable path to reduce their costs by 10-100x over the next 5 years if they stop improving the models. That would make them an extremely profitable company with their only real risk being “local ai”. However customers have w
306.
by lumost 1y ago
That kind of behavior is far less successful than many would expect. Higher ups are pretty aware of who does what, at least at 1 year+ time frames. That being said, I’ve seen many outrun their own incompetence. Getting incentives right in l
307.
by lumost 1y ago
Was the scam that the codec had the raw video in it, so the "files" could be made trivially small?
308.
by lumost 1y ago
Im on an academic foray to an R1 right now. One of the big items the PIs are drilling is the one paper per year mantra. Many papers today are representing what would have happened via open source repos in yesteryear. Meaning that there is a
309.
by lumost 1y ago
We don’t know how sticky that revenue is, or if it’s going to be a commodity in the long run. Similar things used to happen in ad-tech before investors got wise that there was no moat.
310.
by lumost 1y ago
Is it that impractical that the GPUs would find another use case? Of NVidia cut enterprise gpu prices by 10x (which they absolutely could!) the gpu+memory combo would be cheaper than all other compute for database-like operations, simulatio
311.
by lumost 1y ago
The real question would be the debt on the GPUs. If a hyper scalar puts 100B from profits into data centers, and borrows 400B against the datacenters (or their suppliers do!). Then the buildout could be quite problematic. If it’s just profi
312.
by lumost 1y ago
Part of the challenge is that machines are significantly different. The radiologist’s statement that an object measured from two different machines is the same and has not changed in size is in large part judgement. Building a model which c
313.
by lumost 1y ago
There are frames in simulations though! Typically measured as time steps. That the frame usually has N_d dimensions is insignificant.
314.
by lumost 1y ago
I have a thing where I immediately doubt any ML paper that imitates a process then claims that the model is sometimes “even better” than the original process. This almost always means that there is an overzealous experimenter or a PI who di
315.
by lumost 1y ago
I’ve always wondered if it would be cheaper to just have everyone check their bags and eliminate the overhead bin. I wouldn’t be surprised if airline boarding was sped up by 2-3x this way.
316.
by lumost 1y ago
Not to mention the lack of standards on leg room/entertainment packages/food quality for any of the above combinations on any airline!
317.
by lumost 1y ago
That, and all languages more or less converged to the same functionality with different syntax/naming conventions. They also all offer similar toolchains with similar frameworks for nearly all problems. The remaining places where one l
318.
by lumost 1y ago
It's real revenue, but you are operating a fractional reserve revenue operation. If the person your investing in has trouble, or you have trouble - the whole thing falls over very fast.
319.
by lumost 1y ago
Was this all actually an agent? I could see someone making the claim that a security research LLM should always report issues immediately from an ethics standpoint (and in turn acquire more human generated labels of accuracy). To be clear,
320.
by lumost 1y ago
Anecdotally, on green field projects where you are exploring a new domain - it’s an insanely productive experience. On mundane day to day tasks it probably takes more time, but feels like less mental bandwidth. Coding at full throttle is a
321.
by lumost 1y ago
https://wtfhappenedin1971.com/ While you are correct that real wages are up around 25%, productivity has nearly doubled. While various consumer goods, and technology have seen large improvements - ignoring the measurable an
322.
by lumost 1y ago
> However non tech/lawyer/doctors have been adversely effected by the fact they've seen no real income gains in 25 years overall. We may be reaching the breaking point where Americans view any solution to this problem as w
323.
by lumost 1y ago
At the local/hobby scale, it’s very much a “do whatever” area. But I can rent a 4090 for a little under a dollar an hour, and I can rent a b200 for $6, it’s very hard to claim I’ll use 10k+ hours of gpu time on a b2000 I buy for myself
324.
by lumost 1y ago
I don’t really get why anyone would be buying ai compute unless A) to your goal is to rent out the compute B) no vendor can rent you enough compute when you need it C) you have an exotic funding arrangement that makes compute capex cheap an
325.
by lumost 1y ago
Which leads to multiple questions - if NVidia can't manufacture elsewhere, will the best in the industry get much more expensive? is the GPU value in CUDA or TSMC? At best, the two companies are locked into a monopsomy/monopoly
326.
by lumost 1y ago
Nvidia needs more silicon manufacturing partners to prevent TSMC from charging a significant premium. Their choices are very limited today, and almost non-existent outside of Asia. Given a real geopolitical risk and NVidia’s current valuati
327.
by lumost 1y ago
I've become extremely nervous about these sudden declines in quality. Thankfully I don't have a production product using AI (yet), but in my own development experience - the model becoming dramatically dumber suddenly is very diff
328.
by lumost 1y ago
Is the media even “social” anymore? How much of Reddit is simply bots generating catchy takes and then generating commentary on these takes. You can easily be deceived into thinking that a vast number of people believe something, or think t
329.
by lumost 1y ago
It’s interesting that the U.S. picked an employer-driven model, which effectively outsources immigration selection to firms. That’s efficient for demand-matching, but it concentrates bargaining power in ways that a points-based model avoids
330.
by lumost 1y ago
It’s a function of time. For far too many people, the existence of modern life consumes more time than it did a generation ago. We work more hours, we work harder hours, we consume entertainment for more hours. The costs of this societal sh

More ›