It means many more startups 
form by managers and programmers not needed by corporations.
Companies only need middle managers after a certain size, say 20 employees. If one employee can do the job of 10, many tech companies (maybe even most) can stay small and not reach that threshold.
Yeah, there will be lots more startups trying things that previously wasn’t economical to try. And people still say SF is dead.
Bubble forming.
3 Likes
Jensen is playing hardball. The hyperscalers are all developing in-house chips to lessen dependence on Nvidia. And Jensen makes sure they aren’t getting all his chips. Broaden the field and keep these frenemies a little bit on edge.
Nvidia says it gave the startups, CoreWeave and Lambda Labs, GPUs known as H100s , which are up to 9 times faster at AI software training than the previous generation of GPUs. AI developers are falling over backward to get their hands on them. […]
To meet demand for GPUs, CoreWeave plans to triple its number of data centers in the U.S. to nine by the end of the year […]
By including CoreWeave and Lambda in the early rollout of the H100, Nvidia is essentially limiting how many chips AWS, Google and Microsoft can buy—and making sure its end users have options .
What coding tool is he talking about? Let me use it too.
I doubt generative AI can replace programmers and engineers.
Generative AI is very good at “pretending something in fuzzy way”, hence, very good at linguistics and art. However, its weakest point is “can’t judge true or false”. As a matter of fact, it says/produces something false with too high confidence, hence, quite concerning.
Someday, a different form of AI can achieve this as well as human brain, but at least, I don’t see it yet.
From this sense, I think lawyers and judges would stay for a while. (I originally thought that lawyers would be gone soon.) Maybe, we would see new jobs like “fact checkers for AI” soon.
2 Likes
I think you have typo in “Current market cap of AAPL $792B”. Did you mean TSLA? AAPL market cap is 2.9T.
1 Like
.
Yes. Updated.
I saw many good reviews of GitHub copilot. Have you used that before?
Most of my coding is done at workplace and I haven’t seen anyone using AI-based coding suggesting tool yet in my group.
I searched on Copilot and see two potential issues based on above review.
-
A common concern that arises from the usage of GitHub Copilot is whether you will unwittingly copy some open source project’s code. GitHub Copilot has been trained over a large number of open source projects. Therefore, it may theoretically suggest a code snippet from an existing project. Some of these projects could be reserving all copyrights, meaning that you are not allowed to reuse code from them at all.
-
I do not have data on Copilot’s accuracy. According to my anecdotal and likely very biased experience, its recommendations are around 60% accurate. In practice, you eventually get a feeling of when the tool’s suggestion is wrong and learn to ignore it at a glance. However, when the recommendation seems correct, you may end up going through it to realize that it isn’t, or it needs corrections. This could obstruct your flow and depending on your personality and preferences, it may be detrimental to your focus.
However, it seems worthwhile trying at least. If many folks find it useful, there should be a good reason.
3 Likes
Your comments are valid concerns. But it’s still early days. I bet all these tools will improve a lot in the next few years.
1 Like
Looks like another all time high for Nvidia?
2 Likes
My dog isn’t much help as an assistants which means Gates benchmark of a digital assistant isn’t close. Calling this phase generative AI is a bit misleading. It’s a summarizing tool. There’s not even an AI tool that can perform simple tasks. If I’m driving home and want to get food that’s within 1 mile of a highway exit, it can’t provide a list of options then navigate me to one. That’s a pretty simple task.
Think about how many billions around the world drive every day. AI can’t even do that safely yet even with how much money spend on hardware and mapping roads? Humans don’t need to map a road before they can drive it.
One of the original AI stocks that some were super optimistic about. Acquired for $1.7B. That’s a cool 62% discount to its 2019 high.
Sold iRobot long ago… can’t remember whether I have gained or lost money.
There are too many competitors. Btw, I have two iRobots at home, pretty useful… no need to do any more floor cleaning except occasionally cleaning hidden corners and behind doors.
There is a black market of Nvidia chips in China. If you paid double and ask the right guys you can get them.
It takes massive quantities to do meaningful training though. Companies need to spend $5-10M on servers with the chips.
Speaking of GitHub copilot…
1 Like