Unrelated, but is this a case of the Pareto Principle? (Admittedly the first time I'm hearing of it) Wherein 80% of the effect is caused by 20% of the input. Or is this more a case of diminishing returns? Where the initial results were incredible, but each succeeding iteration seems to be more disappointing?
> but each succeeding iteration seems to be more disappointing
This is because the scaling hypothesis (more data and more compute = gains) is plateauing, because all text data is used and compute is reaching diminishing returns for some reason I’m not smart enough to say why, but it is.
So now we're seeing incremental core model advancements, variations and tuning in pre- and post training stages and a ton of applications (agents).
This is good imo. But obviously it’s not good for delusional valuations based exponential growth.
We're seeing diminishing returns in benchmark space, which is partly an artefact of construction, not an absolutely true commentary on how things are progressing.
Well yes but there is no better way to measure without resorting to pure hearsay. How would you make an accurate assessment of something so inherently vague?
Alter the benchmark space that we care about, for example focus only on ARC-AGI-2 and then suddenly the gains are no longer diminishing but are accelerating.
Your comment made me think of the Terry Pratchett book "Going Postal" in which a conman is put in charge of the post office and quickly realizes what you said: selling stamps is free money. One of my favorites from his later discworld books.
It was a car engineered to satisfy California's short lived requirement for auto manufacturers to make a zero emissions vehicle, hence the availability. Despite that, drivers loved it, and didn't want to give them back when GM cancelled the program (after lobbying the law repealed).
The car was engineered with 90s battery tech so it was plagued with problems that anyone who used such batteries in the 90s know existed: high self-discharge, memory effect, low energy density, voltage depression over time, and limited cycle life. These cars wouldn't last 2 years without needing their batteries replaced. That's why they never sold more than 600 of them.
"Memory effect" in the 1990s is an old wives' tale. It's a real condition discovered in the 1980s on satellite platforms with computer-controlled charging, but was identified and fixed quickly.
It never existed in consumer applications of NiCd batteries, especially as late as 1996.
It was not an old wives tail. I worked in wireless retail 23 years ago and saw these problems first hand. NiMH phone batteries from that era would scarcely last two years. Of course, it mattered less then because the tech was improving so rapidly that most people wanted a new phone every 2 years anyways. NiMH was an improvement over NiCd, but it still had memory problems nonetheless in its first few generations (modern NiMH batteries are better at this).
It could be mitigated by fully discharging a battery before recharging, and I'm certain that in applications such as satellites , they engineered the charging cycles to mitigate this. However, consumers powering phones and laptops can't be expected to maintain such discipline. Certainly people driving cars over variable distances can't be expected to uphold such requirements either, out of absolute necessity to travel a fixed distance between charges.
Lithium ion ultimately won because it solved these problems altogether. Modern NiMH has caught up a little bit, but Lithium has meanwhile improved as well.
You realize the https://en.wikipedia.org/wiki/Crawler-transporter is a hybrid EV right? Along with most trains and tanks and ships. But I'm not compensating for anything, so a small vehicle that gets me from A to B quickly, quietly, cheaply, safely, and has a 6ft long bed with fold down sides works great for me.
Those aren't really "sales" numbers (tbh I'm not even sure GM "sold" any car either given that it was able to destroy all of them).
GM only produced 660 units and 457 units. They never tried to actually mass produce the car and seeing how current electric vehicles are I really understand why; why cannibalize your higher margin ICE vehicles?
This is basically the same argument people make about a cure for cancer. Since there's a ton of money in treating cancer you can't develop a cure and kill your cash cow. It however, completely misses that somebody that isn't currently treating cancer can come in and develop a cure (i.e. Tesla) without tanking their treatment margins.
If you're in the U.S. and a "government agent" told you to use iMessage, you are 100% being scammed. No way they would accept anything less secure than a fax message or a document portal that looks like it was set up in the 90's.
I can't speak for Waitrose specifically, but luxury brands will definitely do a comprehensive analysis of the surrounding area before opening a store. In the US at least, the equivalent is looking for the closest Starbucks or Apple store.
Can someone explain. From their definition it looks like the credits and debits column in their example transaction table is flipped e.g. the first "debit" to the cash (debit normal) account increases the value, while the "credit" to the equity (credit normal) account also increases the value. (edited for spelling)
