If you're afraid of spreadsheets in the financial industry, you should probably just keep your eyes closed. The entire industry is built on them to a very large degree. I've been in IT for almost 30 years, and over most of that, I've seen Excel spreadsheets used throughout the various accounting departments of companies ranging from small operations of a couple of dozen people up to multinationals with tens of thousands of employees, with some banks in there. Some of those spreadsheets are enormous and are doing complex calculations across sometimes a couple dozen worksheets, where changing one number can take several seconds to recalculate everything even on a reasonably modern computer.

Uh, some of us in Windows land use PS/2 for NKRO gaming.

Then they'll find out the hard way that Steve was right.

and it's here today. No homelessness. No medical bankruptcy. No food insecurity. Affordable education. Decent places for everyone to live.

Good movies have both.

Hard disagree. There's absolutely nothing preventing a fun movie from having decent writing (i.e., "treat your audience with some basic respect") aside from cheap studios and hack producers. Michael Bay and JJ Abrams should have been warnings, not instruction manuals.

Sometimes critics focus on silly or tangential things in movies that average people don't care as much about, but critics are also much less willing to let incoherent plots and repeated non-sequiturs pass just because CGI and EXPLOSION.

Ironically, Altman agrees that AI is a bubble:

That's a lot of word salad from someone who has never manufactured a solar panel or used one.

Sit down, son, a real semiconductor professional is speaking.

Solar power is free. Hydrogen makes for fairly decent storage once you solve the 'bleeds through most anything' part.

Hell, I get a refund check from my municipal power company every year, right here in California. Do you?

I haven't seen a power loss unless it involved a vehicle smashing into something that affects power distribution. I've been here nearly 2 decades.

Unlock all interactions? No. Unlocking a specific interaction? Maybe.

For common uses (like the public web), the most likely approach to decrypting a specific interaction is to break the RSA (cert-based) on the outside and then the Diffie-Hellman (ephemeral per-transaction) on the inside, then recover the symmetric encryption key to decrypt the rest of the conversation. But this is not trivial, and it requires more work than to just toss the transaction into the quantum computer.

The ephemeral layer is where things get harder. Even if you can derive the RSA key on a regular connection, you've got the first layer, but the DH layer is redone for each new connection. (Some sites don't use DH, or are vulnerable to downgrade attacks where DH isn't used, but DH is pretty widespread.) Every ephemeral negotiation has to be individually cracked. Tor uses DH or x25519 on all connections, so each has to be individually cracked. It is expected that breaking an individual 2048-bit RSA or DH encryption would take several hours if one had a quantum computer of sufficient power. Cracking 3072- or 4096-bit RSA/DH will take even longer, if it's even possible on the same systems. However, we appear to be a long way from such capabilities, and the NSA isn't likely to use it to break arbitrary Tor connection encryption, saving it instead for much more practical items. As soon as the NSA has practical quantum computing, it's going to have decades of backlog to go through just for the international signals, and getting anything moved up in line is going to need a damned good reason.

> Making Alexa is not THAAAAT hard

Last year, the WSJ reported that Amazon lost $25 billion on Alexa from 2017-2021, partially from selling devices below cost but mostly because of development costs. It seems like it's harder than you think.

Not just once the crash comes, but this sort of thing will help cause the crash.

Cory Doctorow talked about some of it in a recent post that really helped me grok what I've been feeling about the inevitable AI bubble collapse.

AMD giving 10% of itself to OpenAI is almost entirely symbolic as OpenAI will turn around and "spend" that money on something like GPU hardware, probably AMD products. AMD is just paying itself via OpenAI and both companies will book it as revenue to try and obscure the colossal losses on AI spending.

It's all monopoly money laundering, and it's going to crash hard.

Amanda Langowski's family information.

Seems only fair to return the privacy violation in kind.

Nvidia is therefore a bubble. This article is complaining that Europe is an obstacle to further bubble inflation.
No amount of Nvidia etching IP onto wafers is worthy of a 4.6 TRILLION market cap - bigger than the 4.2 Trillion market cap of the ENTIRE name-brand pharmaceutical industry.

Correct.

If you are not a tech person, right now from all the hype and news and bullshit, you would rightly assume that AI is an amazing revolution and you would barely have heard about its shortcomings.