Hacker News Reader

Comment by abrichr | Hacker News Reader

Serious question: is there any reason to trust anything IBM does or says isn’t vaporware? In recent memory I am only familiar with failed projects that received copious media praise (eg Watson, BlueMix).

The only project I know of that was genuinely successful is Deep Blue. What successes have they had since then?

View on HN

IBM makes an enormous amount of non-vaporware, you just never hear about it because the licensing situation for most of it is so atrocious and its probably exclusive to their system z (mainframe) architecture. IBM systems is a top tier organization being held back by incredibly restrictive business practices. I'm lucky to have been in positions to see a lot of what they work on, but it's always tough to explain to folks who's most memorable recent exposure to IBM products is likely to be some sort of PR stunt like Watson. I've held out hope for a political change within the organization in favor of the systems folks for a long time, but any success on that front has been pretty limited so far.

Kinda-Sorta. It depends on what you mean by "success". IBM has been doing a commendable job at managing expectations wrt/ quantum computing. e.g. from this very article:

"The new 433 qubit 'Osprey' processor brings us a step closer to the point where quantum computers will be used to tackle previously unsolvable problems,"

So we can reliably expect them to ship what they are promissing. However, there is still a massive skew between what QC brings to the table vs what the public (and investors) are expecting from it. At the current trend, anything QC-related is effectively guaranteed to qualify as vaporwave due to the public at large being mislead on the subject for so long.

This processor has the potential to run complex quantum computations well beyond the computational capability of any classical computer.

I don't think that manages expectations well, the word "potential" is doing a lot of work.

Although I'm not up to date on the most recent industry developments, I'd say to be less concerned about the number of qubits, and more concerned about the gate fidelity and qubit coherence. Certainly, you want a quantum computer that you can actually use [1]! It's also important to distinguish between the adiabatic quantum computing of companies like D-Wave with their many qubits and the gate-based quantum computing of Google, IonQ, IBM, etc. with their fewer qubits.

[1] - https://arxiv.org/abs/2110.03137

Those flagship projects that failed technologically were amazing successes.

> received copious media praise

That's the success criteria.

IBM makes money in lots of ways. The last few decades they've struggled with public perception, and an image of being "has-been". This has hurt their ability to get talent and enter new markets. Having your name next to neat, futuristic tech is a good way to counter that image. Even more so when it's things like Watson and Quantum computing that have real nerds doing actual fawning over the tech.

There is successful tech in the “tech that does what says on the box” and in “achieves good market fit”.

A lot of their stuff does 1, but not 2.

And their mainframes, underhyped to the extreme, that more or less still run the world.

I didn’t know IBM mainframes run the world? Just curious, care to expand?

Banks run the world, IBM mainframes run the banks.

>IBM mainframes run the banks.

And the airlines, and the logistics companies

That's a bit unfair to the insurance companies. They have a pretty big say in what happens and also run mainframes.

And mainframes run on concrete floors, maybe construction companies run the world by that logic?

Mainframes aren't a big truck, they're a series of concrete floors.

My usual comparison is with cargo airplanes.

This is becoming less true over time.

many of the most core US IT infrastructure projects- government and commercial- depend on IBM 360 assembly code or another IBM system, written decades ago that has been repeatedly scaled to handle the increase in load.

Two I can think of: SABRE and the IRS. See more here: https://www.nextgov.com/cxo-briefing/2016/05/10-oldest-it-sy...

Personally I have a ton of respect for production systems that can be scaled in place for 50 years. But I don't have any interest in learning to run IBM mainframes or code for them- I'm preparing for lots of contract work in 2037.

> And their mainframes, underhyped to the extreme, that more or less still run the world.

More or less, but less and less. Many a bank, airline, etc. has had or has or will have a "digital transformation"/move to cloud/rewrite apps/websites to actually be usable/etc. project to switch away from legacy stuff handicapping them to a more modern leaner tech stack. Mostly because they got shown by newer, leaner competition how it's done (neobanks, low-cost airlines) so finally there is some incentive. And of course the added benefits of commodity hardware (as a Service) being much more easily replaceable and scalable. If you're a big bank/airline, why would you spend billions on mainframes from a single vendor you're beholden to, with all the staffing problems (nobody knows this stuff so hiring is a nightmare), specced for the highest possible load you might get? Instead of throwing it on AWS which "everyone" knows and can scale up/down based on demand. Migrating isn't easy or cheap, but it makes so much sense.

Power architecture chips, but that's from a no-need-for-marketing-embellishments classical realm.

https://en.wikipedia.org/wiki/Power10

for an example that is current

Was Watson a scam? I mean the jeapordy AI, not all the completely unrelated projects they gave the same name later.

Consider that the Jeopardy stunt was basically a sales pitch saying: "If it can do that for Jeopardy, then just imagine what it can do for you!"

There is definitely an argument to be made that this was an empty promise. Not necessarily enough to call it a "scam", but vaporware...

I don't think it was an outright scam, but its use in medical analysis turned out to be very poor: https://qz.com/2129025/where-did-ibm-go-wrong-with-watson-he...

Definitely over-hyped technology. Probably not a scam though.

In light of the large language models from Vaswani et al., I'm curious if anyone could chime in:

1. Does Watson incorporate any deep learning?

2. If not, why? Were disagreements based in business or theory?

"If IBM Watson can find hidden correlations that help your business, then why can't IBM Watson stem a 3 year sales drop at IBM?" - random Twitter user

Kasparov was disappointed with Deep Blue. After the rematch won by Deep Blue (the first match was won by Kasparov), IBM relocated the machine then disassembled everything.

Reply on news.ycombinator.com