> Also known as a monster-in-the-middle,[1][2] machine-in-the-middle,[3] meddler-in-the-middle,[4] manipulator-in-the-middle,[5][6] person-in-the-middle[7] (PITM), or adversary-in-the-middle[8] (AITM) attack.
> You are not a person or entity that is: (a) located in, organized under the laws of, or ordinarily resident in any country or territory that is the target
of comprehensive U.S. sanctions; (b) a prohibited or restricted party under U.S. or other applicable sanctions and export control laws and regulations;
or (c) owned or controlled by or acting on behalf of anyone described in (a) or (b). You agree to use Let’s Encrypt Certificates and any services provided by or on behalf of ISRG in compliance with applicable U.S. export control and sanctions laws and regulations.
This is about inner product functions in a specific kind of Hilbert spaces, a notion that is very useful in many branches of applied mathematics. Machine learning and functional analysis included.
The name collision is unfortunate.
One unfortunate difficulty is that these kernels don't map so well to GPU kernels unless explicitly embedded in high dimensional spaces. This is one of the reasons why kernel methods has recently fallen out of favour in machine learning - lack of mechanical sympathy. Note this just one reason, there are others.
Most of the GPUs built in the past year are sitting in warehouses, just waiting for data centers to break ground and for electrical expansions to complete. And Nvidia can’t get China interested in buying their products in the Trump era.
There will be a massive glut of hardware soon enough. OpenAI needs $532billion in cashflow in the next 4 years to keep the “infinite money glitch” going. That’s not likely to happen unless AI makes some 10x value improvements for their customers in the next 1-2 years that we aren’t seeing now.
LLMs are terrible at generating code for “less commonly used languages”. They require LOTS of data for high accuracy.
I describe it this way: they are good at interpolating from what data they were trained on, but terrible at extrapolating. I agree with the parent that the LLM-generated content isn’t novel, it’s just a rehash of two things it was trained on.
I have wasted quite a number of hours trying to use LLMs to write things for less common languages. Sure they can one-shot some impressive stuff in C#, Python, and JavaScript… but try working in Object Pascal: it’s non-obvious hallucination after non-obvious hallucination, presented confidently enough to make it difficult to see it’s complete garbage, so you waste a ton of time trying to polish a turd.
Creating a new paradigm is a problem with a lot more groundwork laid that working in an existing little-known paradigm. One is creating patterns which only have to be good to be correct. The other has to be correct to be good. They are completely different problems.
I’m less anti-“anything AI” and more anti-“how AI seems to be used right now”.
It is:
being used as smokescreen for massive layoffs industry wide.
a repeat of the business models of 1999-2000 (growth without profit, race to IPO, promises of infinite TAM).
the business execs are following the crowd, insisting on token maxxxing, not value-to-the-customer maxxxing. There are reports that many companies have already exhausted their annual AI budget by April.
most companies don’t know how to measure if it’s actually increasing value AT ALL.
my former coworkers say it’s empowering non-engineers to write bad code / features which are a net negative. Prior to AI, bad ideas needed an engineer to tacitly approve of it, but now the bad ideas can bypass the engineer at light speed. 10x development speed is VERY BAD if the average change is a net negative.
people are leaning on LLM inference instead of basic tasks such as keeping web bookmarks organized. It will cause cognitive atrophy.
the foundation model companies are heavily subsidizing my $20/mo plan, so I’m pretty sure it becomes unaffordable once they charge cost plus for inference.
I’m personally experimenting with AI for all aspects of business (not just coding). I’m sometimes vibe coding prototypes and sometimes building a rigorous full-SDLC app.
I watched an interview with Ed Zitron yesterday and I found myself agreeing vehemently with his appropriate level of cynicism about the AI industry and how business is using it right now.
Localhost is “on the device itself”, but so is an installed App and files and user settings.
This is also missing a lot of what localhost means in this context (networking, violation of the usually way similar apps and websites work on an Android device, etc).
The problem is the stock market is more divorced from reality than we have ever seen. For instance, why does Tesla stock still sit where it is? How could it possibly not be going down at this point? So many undelivered promises, major setbacks in sales, massive decreases to their sales forecasts… literally nothing has gone well for them in years and yet the price is still outrageous. It really feels like I’m just out of the loop on something.
Jack Barker’s rather blunt monologue in SV about how the stock is the product is more true than ever. It felt very heavy handed at the time but it’s only proven to be more the case than I thought.
This probably illustrates my disconnect from reality, but I’ve never understood why a company would care about share price once they’ve left the door. I get that the co still owns its own shares and can conjure new ones for sale, but why would those very infrequent events interfere with the day-to-day operations. In my (wrong) eyes, it’s like pro-baseball players trying to increase the value of their trading cards via their participation in the game. The team doesn’t matter any more, it’s al about what the card owner wants.
Company itself really shouldn't. Everyone involved in management from board to executives do. Board operates behest of stock owners, executives operate behest of board. Such to keep their job they have to do what stock owners want. And stock owners either want dividends or growth in some term.
> Board operates behest of stock owners, executives operate behest of board
These are often both weak signals, though. They'll govern very high level decisions, but all the day to day is inside the company. Just as I want a return on the money in my bank account (as I was promised) investors want a return on their money too, and as you say, the executives and board should care about making sure the people who put money into the company are getting a decent deal out of the arrangement.
People have always way overstated the power and scope of “fiduciary duty.” It doesn’t mean you have to redline your company at all times to maximize every single penny in the short term at the expense of all other considerations. That’s just a cultural thing we do in the US by choice
Selling more stock is usually a lever a company can pull when they want. So even if a normal company in normal times doesn't have a reason to do so often, they can if circumstances change. Tesla and some other meme stocks have been extremely aggressive about selling more shares into crazy valuations, and have raised immense amounts of money doing so.
Plus as others have said, usually all of the decision makers have a bunch of stock exposure and will prioritize their own financial gains over pretty much anything else.
I used it. It doesn’t work as FSD. A driver has to pay attention and intervene. Can’t sleep. Can’t read a book. Can’t look at the scenery going by. It’s still super neat and I like it. But it’s not FSD, and I suspect why fewer than 10% of tesla owners pay $100/month or bought it.
> Also known as a monster-in-the-middle,[1][2] machine-in-the-middle,[3] meddler-in-the-middle,[4] manipulator-in-the-middle,[5][6] person-in-the-middle[7] (PITM), or adversary-in-the-middle[8] (AITM) attack.
reply