Something that makes it easier to own EVs here in London (UK) are these chargers on lampposts, there are probably thousands of them at this point: https://i.imgur.com/1YdeVwf.png
They're slow (3-7kW), can't be installed everywhere, there are not enough if everyone wants to charge at the same time, and some areas still don't have them, but plug it in when you arrive and you should be fine to go in the morning. It's a good idea and more places should adopt it.
Not just in cities. In the U.S. at least, a lot of people live in suburban townhouse complexes that have shared parking lots. It's going to take a lot of investment to make charging easy for everyone.
That wouldn’t be a problem for the op with the phev because they are already charging at home and only driving less than 30 miles each day except 8 times in the past year. I still don’t get why phevs are a thing.
My highly opinionated take from working in a field with replication issues -- the people publishing unreproducible results simply want to either establish or reinforce high social standing.
They are highly intelligent and skilled in the sense that they can progress their career through complex political moves within funding agencies, journal editorial boards, conference organization, and university departments. Proper statistical analysis and experimental design are absent because it's a nuisance in the way of success, not due to lack of understanding or low intelligence. There's still room for rigorous scientists to succeed, but it's becoming untenable for many to stay.
It depends, the compensation is highly skewed. The PI in question is likely compensated over $1M/year, while other investigators involved (assistant and associate professors) have compensation comparable to entry-level programmers. To confirm these things, you can explore this database (only public CA universities though): https://www.sacbee.com/news/databases/state-pay/
$1M/year sounds insane. In Finland at least top salaries for professors is about tenth of that. And I don't think any of the professors I know make much more, in Europe or North America.
Admin people like the principal may get obscene salaries (like 4 times the max professor salary) though.
As an academic, I think academics (including me) are paid too much. We need people who are interested in the science, not those who (pretend to) do it for money.
> We need people who are interested in the science, not those who (pretend to) do it for money.
In the US, the problem is mostly the opposite. Lots of people would rather do science, but it's hard to choose using your skills for real science when some company will pay you 2x-10x to instead optimize ad clicks on their website or algorithmic trading or whatever.
In some of the humanities, liberal arts, social sciences etc academia may pay better than other options for those people. But for most STEM folks, it's a tradeoff between "low-paid meaningful work" vs "high-paid meaningless work".
Medicine and AI research may be the only two areas where people can simultaneously do cutting edge research and make high salaries.
Maybe a bit harsh, but I'm not sure science would get more trustworthy if we'd get more people who choose to optimize clicks just because it pays better.
It's really easy to do "shortcuts" in academia that gets one better salary if that's what they want. As seen in this post.
It's rare for an academic to make over $1M in salary, but mid-high six figures is not uncommon for a highly published, well known name with a track record of winning large grants.
While that seems reasonable, it assumes the amount of impact per paper has remained constant, while is practice it has greatly declined. These days scientists will literally try to identify the set of "least publishable units" in a research project to maximize the number of papers (I've attended meetings where this is discussed without irony). There is also a tendency to publish work that is quite similar to previous work with a minor change that is emphasized disproportionately to its importance.
Counterpoint, in the context of education, the goal of the assignment isn’t to produce the end product of a 3000 word essay, but to have you go through the associated process of formulating, organizing, and presenting your thoughts. LLMs give you the product, but not that process. Outside of school the product matters more though, so learning to use LLMs as a tool seems like a worthwhile part of education as well, as long as you learn enough about writing to prompt and judge the quality of its outputs.
Yeah. A university paper assignment and a 1000 word article for the company blog seem like two totally different things even barring explicit policies. (Though I'd be surprised if a lot of universities didn't have policies by now.)
I understand that, and I go through the process. Even though we know that " the goal of the assignment isn’t to produce the end product of a 3000 word essay", it doesn't change the fact that I have to produce a document with 3000 words that could be written and summarized in 1000. IMO, with such tools, we should maybe question if such "Write an essay with X words" still make sense.. I'm sure the (instructors, master and doctors) that do the review of 3000 word texts would welcome such change.
Beyond this, some reviewers use their position to take advantage of the process to suppress, slow, or preempt work that overlaps or disagrees with their own. The journal editors are supposed to police this kind of thing, but they are similarly providing a community service that is lacking quality.
> slow, or preempt work that overlaps or disagrees with their own.
No doubt - past two papers I've submitted I can basically deanonymize my reviewers based on the other papers/projects they insist my work is inferior to.
One difference is that humans are actively involved in data collection, so when there is a gap in their knowledge, they don’t just wait for the information to show up, they ask a question, etc.
Much of that time also includes physical interaction with the world, which makes it far more valuable because it can improve performance in a focused way.
FWIW, I took “GPUs are deterministic” to mean they are deterministic in all possible intended use cases. This is not strictly true, since the whole point of using them is massive parallelism, which brings along non-determinism, for reasons that others have noted. Of course it’s possible to choose to forego that, but what is the point of a GPU in that case?
This is a false dichotomy. You can have massive parallelism and determinism.
You can trade determinism for convenience, but that doesn't make things easier: now you have to deal with the determinism.
But to suggest that massive parallelism somehow implies non-determinism is quite disingenuous from my perspective.
We have mutexes and lock-free ring buffers and stable sorts and all sorts of bells and whistles to make parallelism safe elsewhere. We also already have tools to solve this for GPUs.
It does seem likely we’ll soon have cheap enough LLM inference to displace traditional NLP entirely, although not quite yet.
reply