I like the fact that more people can produce demo software now. It means they can produce a rough draft that a professional can use as a starting point to build properly.

Academic writing is vulnerable to opinion masked as fact [1]. Most people do not read the cited text [2,3] so authors can make broad, opinionated statements that seem to be backed by a peer-reviewed literature [4-6]. LLM have just learned to master an old trick.[1,2,5] [1] ref that is not relevant [2] my own work on a different topic [3] retracted paper [4] my friend's paper [5] high profile review article [6] a paper that disagrees with this point

This is what I have observed in Canada as well. Positions are being filled by people with surprisingly few technical skills. This isn't to say that there aren't technical people in government, but it seems like the hiring process doesn't adequately test for core technical competencies.

If you can't keep a normal boat operational, good luck with one that is full of sensors, cameras, computers, etc.

Given the impact that neural networks have had on physics research in the past few years, I can see the logic behind this.

Since when are LLM leaderboards prestigious? And is anyone surprised there are new models coming out with larger context lengths? I wouldn't call that a shockwave through the tech community...

The article makes it sound like a PhD in CS makes you better suited to work with data than an undergrad in physics. This is does not agree with what I've seen. I'm sure there is a joke to be made about sample bias and anecdotal evidence, but notwithstanding that, this is what I have seen.

My understanding is the the batteries for EV are no replaceable without great cost. How can I be expected to buy a car that will be worthless in 7 years?

This is rules based. This isn't going to be particularly effective in preventing fraud.

Alexa has been terrible for years. I bought the initial device when it first came out and continued to buy hardware when it was on sale during Prime Day, Black Friday, etc. I was so hopeful that it was going to be a new way to interact with technology. What a disappointment. For example, I listen to exactly two radio stations: CBC Radio One Ottawa, and KQED. Invariably, when I say "Alex, play CBC Radio One Ottawa", it will mishear me and launch a radio station that has a very dissimilar name. It is clear that the backend system translates the audio-to-text, and then does a lookup. It completely ignores historical queries - this means that the audio-to-text step does not get simpler with time (as it should). They are clearly not using a ranking algorithm on the responses either, otherwise my common choices would be very highly ranked. They also never solved the key problem with shopping - I don't trust Amazon to buy products that are reasonably priced. All they needed to do was provide a simple web-ui where I could put some guardrails around price for common items and I would have used it. Alexa was great hardware that was well ahead of it's time that was totally mismanaged on the backend. The fact that they haven't launched a beta chatGPT version just shows this mismanagement continues.

Yann LeCun has been railing against chatGPT enough that I suspect he is getting a lot of pressure from inside META about why they didn't launch something like this. The sad thing is that they DID launch https://galactica.org/ but then pulled access to it almost immediately after.

How is Tesla a top selling EV brand when their cars have so many issues? I don't get it.

How is it possible that Julia is so fast compared to Python when both are apparently interpreted languages?

This device never made any sense when we are all walking around with phones in our pockets.

This seems like it would be best paired with old text written by the relative. It would be their thoughts, just played to you in a different medium. That seems much less creepy