I, For One, Welcome Our New Self-Driving Overlords

@BrianRoemmele: Phone confessions

I had a long talk with an executive at a large AI company about the realities they face. He was one of the folks that booked cohort 1 of my 17 minutes with me, it was for non corporate folks, but he was honest with me and paid 17 times (no accident, thank you) more than the minimum.

He wanted to remain anonymous but I told him I knew who he was but will ignore it. I had my plan beforehand and built an approach, honesty my only poach.

My 17 minutes of monologue (really non stop talking) was centered around on this patent and the story of how a massive well funded company had the future that would take it out being built literally in THE BASEMENT by a hacker in 1975.

Takeaway line: Egos will be very high until the accountants take over.

Update: Read the comments.

6 Replies to “I, For One, Welcome Our New Self-Driving Overlords”

  1. “Egos will be very high until the accountants take over.”

    If anyone is wondering wtf this guy is talking about, it is two things.

    First, there’s a YouTube by musician Rick Beato, fairly recent, in which he predicts that “AI” will go the same route as the big recording studios, and for the same reason. That reason is consumer-grade hardware. He demonstrates that a -boomer-musician- can download and run a perfectly functional LLM agent on a PC. Advantage, the LLM can’t phone home to Big Company and give them all your work for free. 15 minute video.

    That means that all the incredible cubic billions of dollars being spent on compute servers is going to be obsolete in 3 to 8 years. Just like big recording studios that cost millions to construct were replaced by AbletonLive and a frickin’ laptop. Whole recording studio for less money than a Gibson Les Paul, just add talent.

    Second, and maybe worse, is that the Big Iron approach is being taken because they’re all trying to brute-force their way to victory. The belief is that throwing enough training data and compute hardware at an LLM will push it across the scaling law boundary into fabulousness. Thus far, this is proving to be untrue, and the scaling law is a hard wall. To get a 1% increase in LLM performance you have to throw 100% more compute resources at it. Or maybe 1000% more for all I know, I’m not an engineer. Hence the data centers.

    Don’t get sucked into the hype.

    1. Should be interesting to see what happens to the AI data centers. Do they switch to cloud, crashing those rates? Do they shut down and sell off the gpus to basement dwelling teens? Do the mine Bitcoin?

  2. Well this genius in the basement (or garages or kitchen tables) are never encouraged in ANY corporate environment especially “go to market” startups…

    I chuckle every now and then thinking about how some time ago my brother thought of and built a device on his own time, for detecting and measuring the PPM of Chlorine in a certain gas the name of which escapes me now. A device the corporation was ultimately thrilled with only because they were stymied by this particular problem for a number of years.
    $$$$$$ were dancing in his head until the corporation reminded him of “conditions of employment” along with a footnote detailing intellectual property and all that good stuff.
    Basically a “Well done, skippy! We’ll take that off your hands, thanks.”

    Anyways. Carry on.

  3. AI is a Craptastic bundle of hubris, promotion and excess.
    Signs of the real apocalypse? Sam Altman begging to Trump for gubmint funding.
    The math doesn’t math, for the cycle of design/build/investment/ rinse/repeat.
    That’s been pointed out by many surrounding that business.
    The depreciation of the asset cycle (the chips) just doesn’t fit the investment required, nor, the lack of profit already embedded in the cycle.
    Power consumption and costs are off the charts.
    Maybe some massive entity can continue in some form or function, but, they are vested in their interests, of course.
    Look at FB/Meta, and their $80 BILLION writeoff on their Alt Universe foray.
    Musk has similar, deep pockets.
    Do the rest? And is that any guarantee of success? No, note even close.
    This has yet to fully play out.
    Someone will succeed, but won’t be obvious at this time.

    1. also holland tulip bulbs, beanie babies, florida real estate, dot com . . . . yawn . . . . . zzzzzzz

  4. from the link:
    “The only company that is facing this is XAI and Elon,
    he just about burned down the entire process. ”
    there it is again. Elon is *different*. his perspective and powers of observation and capacity to NOT get swept away with the hype and crowdthink. this matched by his obvious engineering skills AND ability to delegate to others of the same level.
    does he get everything right? hell no. thus valid but trivial criticism. and a lot of it baseless.

    and he decisively has proven forever that we autistics are capable of great things, just
    get the h-e–hockey sticks OUT OF THE WAY

Navigation