Hacker News Reader

Comment by stathibus | Hacker News Reader

IBM is famous for charging people for the privilege of talking to them, even if you're trying to sell them something.

This strategy makes sense if you consider that even in it's heyday Watson was 95% data science consulting firm and 5% actual valuable technology.

I really think Watson is one of the biggest tech marketing bamboozles of the 21st century. Through Jeopardy they really had a segment of the business world and the general public convinced that they had cracked AI, but behind the scenes it was all one-off custom solutions under one trademark.

View on HN

Deep Blue preceded this. They had specialty equipment for crunching chess moves. Really excellent hype machine.

I liked the documentary about it: https://m.imdb.com/title/tt0379296/

I don't see the problem here, that was literally the point. To build a chess computer.

I think maybe the common problem is that there’s a hidden and false assumption that being good in one domain will transfer elsewhere, it doesn’t and didn’t

That was the last time IBM was cool.

That and it was cool when Watson won at Jeopardy.

I think the above poster's point was that they think Watson's point was to build a Jeopardy machine (and hype).

The Jeopardy thing was pretty brilliant in a sense - wrangling an episode of free advertising from one of the most popular game shows. A memorable episode at that!

Alternatively, there's a great Deep Blue documentary by Fredrik Knudsen on YouTube

https://youtu.be/HwF229U2ba8

Was Deep Blue/Watson basically the starting point of AI/ML, or had that already been happening earlier or concurrently?

AI/ML has somewhat continuous research going all the way back to the First AI/ML Boom of the 1960s and the Second AI/ML Boom era of the "Lisp Machines" [1] in the 1980s. Depending on who you ask the current AI/ML Boom is something like third or fourth wave, and yeah some of that depends on where you fit Deep Blue and/or Watson in the timelines (or even if you bother at all to count them, given all the other comments here on how Deep Blue/Watson have always been PR maneuvers more strongly than technical lines in the sand).

[1] https://en.m.wikipedia.org/wiki/Lisp_machine

Deep Blue isn't ML at all, it's just a purpose built system for solving chess. It doesn't learn anything as such and just a fancy brute forcing machine that (smartly) goes through all possible chess moves and selects the best possible tree of outcomes. The work on actual ML happened independently of this.

That is a form of machine learning. It uses an algorithm with an enormous ability to look ahead, and selects the moves according to a heuristic that have been shown previously to be most likely to create the desired end state.

Chess is not a solved game from the starting state, so it has to make assumptions based on the data it has. This is machine learning. Please don't make such definite statements. You even say it in your own comment "selects the best possible true of outcomes", how do you think it selects this? It uses heuristics to assign values to different board states, and in the case of Deep Blue these values were created through previous game analysis. If a knight to c2 on turn 8 is rarely seen in the same game as a winning board state, then this is valued lower. Looking through the tree wasn't the ML part, but knowing how to pick the best node on the tree was. Deep Blue is ML.

> It uses heuristics to assign values to different board states, and in the case of Deep Blue these values were created through previous game analysis.

Unless I’m misinformed, this part isn’t true. The heuristic was hardcoded with the help of human experts.

From the paper Deep Blue by Murray Campbell et al (people who worked on Deep Blue) "The initialization of the feature values is done by the “evaluation function generator”, a sub-program which was run on the master node of SP system." Which would suggest it generated the heuristic itself. The features may have been hardcoded, but assigning values to them wasn't. In addition to this, feature values could be static or dynamic, meaning it would update dynamic ones depending on the board state. It not only generated the heuristic values (feature values) it could modify them to reflect their relative change in impact throughout the game.

mehh feels like we are creating fancy ML brute force crunching machines now

Now = last 40 years

Deep Blue and Watson aren’t the same thing. Deep Blue was a chess computer from the 90s — and it was not the first chess computer, so regardless of whether you consider computer chess to be part of AI, the answer is no.

It's a shame we never got a cross of Deep Blue and Battle Chess. Make those chess masters fight for their lives.

In fact, forget the chessboards.

Just make them chessmasters square it off in the ring.

Chessmaster Deathmatch Season 1 Episode 1: Kasparov vs. Fisher

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

They bamboozled no one who knew anything. Even the "sold off in parts" bit in the OP article title doesn't surprise anyone who knows what "Watson Health" means. It was never a platform just a bunch of disjoint garbage components. You couldn't sell the platform so now they're selling off the tools. Imagine a toolbox with a spade a fork lift and a soldering iron.

Watson is one the greatest marketing achievements of the 21st century.

It’s literally consulting with open source software. But they packaged it up in an attractive format that got them buzz. Imagine what that marketing team could do for a product that was able to half work!!!

To be fair, paying them for the privilege to sell them something makes more sense than the inverse.

Kindah like Palantir

Ehh...I get why tech people are suspect of Palantir but they're just disrupting other government contractors.

And from my limited experience working with the government, they absolutely need / want / rely on having companies hold their hand as they insist on doing things the hard, slow, and very custom way.

I feel like with this comment, the word "disrupting" has officially jumped shark and lost any and all meaning.

Palantir isn't "disrupting" government contractors. It is a government contractor, țhe old-fashioned kind, at that, and nothing more.

And people aren't suspect of Palantir. People dislike Palantir because its government contracts are shady and boring (they're all about efficiently tracking people), but they pretend to be a "saving the world with tech" startup with such enthusiasm you'd think they put the Kool-Aid into the water coolers.

Unsurprisingly, XXI century Stasi trying to look hip to recruit the tech talent give off nothing but "how do you do, fellow kids" kind of vibes.

Disclaimer: interviewed for Palantir. They thought I could be a fit. Must have not been using their own software to vet candidates back then.

To be honest, my experience of private enterprise is that they insist on doing things hard, slow and very custom. Almost every problem that exists is distinct enough that you can argue it doesn't fit the existing COTS software. It's sometimes necesary to build something yourself, but no where near as often as it is done.

If I remember correctly, John Ralston Saul made the point [0] that there is little observed difference in terms of the efficiency of decision making between large government and large corporations.

My personal experience is that large organisations and government are barely distinguishable. This is often excused in the name of "risk mitigation" - but in my experience it's really just that there are more snouts in the money trough, and that politics is more important than success. The close ties between government and the largest organisations also invite a similar culture.

The end result for both government and large enterprise is "hard, slow and very custom". (Of course, there are exceptions in both government and enterprise).

[0] http://www.johnralstonsaul.com/non-fiction-books/the-unconsc...

This is been my anecdotal experience.

In the UK, I note that politicians extolling the efficiency of the private sector rarely have any substantial experience in big companies.

They have plenty of experience being wined and dined by executives of big companies though.

Some might do it under the belief that it makes them more flexible later on, but going with off the shelf stuff means it's easier to find people to create and migrate to a custom system when or if it's needed.

Disrupting? Their stock price has dropped by two thirds from its peak. They're probably looking to find a buyer at this point, such as IBM itself, while the leadership uses them as a stock-printing machine to enrich themselves.

It doesn't take much to disrupt a fax machine.

Doesn’t this apply to most startups though? The core technical problems aren’t “hard” it’s that the industry involve can’t adapt due to inertia of entrenched companies.

Because of Thiel's lack of ethics and his support for the political far right?

The company is like a perfect storm of bad actors from horseshoe theory; The actual day to day CEO of palantir is Marxist theorist.

>Karp has described himself as a socialist and a progressive, and said he voted for Hillary Clinton. In 2017, he was recorded during a Palantir company meeting claiming he turned down an invitation from President Trump, saying “I respect nothing about the dude.”

>He has said that technology companies like Palantir have an obligation to support the U.S. military. He has defended Palantir's contract with U.S. Immigration and Customs Enforcement during the controversy over family separations, saying that while separations are "a really tough, complex, jarring moral issue," he favors "a fair but rigorous immigration policy". He has said the U.S. government should have a strong hand in tech regulation and that western countries should dominate AI research.

>Karp founded the London-based money management firm Caedmon Group.

>In 2004, along with Peter Thiel (who had been a classmate at Stanford) and others, he co-founded Palantir as CEO.

>He is described as a wellness fanatic who swims, skis cross country, practices Qigong meditation and martial arts, and keeps Tai Chi swords in his offices.

He'd make for a good Silicon Valley or Black Mirror character.

If it was just government...

Can you expand?

It’s promoted as a complete system, but you end up paying them to build out the system to do whatever you wanted to do.

IBM, Oracle, Panatir....in a lot of cases these are pro services companies that custom build whatever is needed. LOTS of money in enterprise application development.

Oracle is a expensive and "popular" database, IBM does many things ( eg. Quantum computers, ... )

Palantir is a consultancy body shop with a dashboard product and a probably good pipeline for merging data. ( Reference - https://news.ycombinator.com/item?id=11646587 )

Pure consultancy is rated much lower ( eg. Thoughtworks if you want a reference)

Consultancy + adaptable software is a decent business model. Unfortunately "adaptable software" for {insert industry} is a really hard target to architect right.

Especially when the majority of your tech headcount bills by the hour and gets paid to tell the customer "Yes."

I didn't say it isn't a decent business model.

In the context of pltr, I'm saying it's not similar as Oracle/ibm and comparison is futile.

That's how it is for just about any non-trivial business system. They're sold as complete systems, but there's still customization required for those systems to actually fit the business - and it's through offering consulting for those customizations that enterprise software vendors make the big bucks.

Right, AI is just custom software, with people in the back to deal with edge cases?

It's IF statements all the way down, baby.

IF (intelligence.getAnswer("What is the meaning of life?")) {...}

Yeah, that is patently untrue.

How so?

They are just a contracting company making money off billing the government hourly for their employees time. It’s a horrible business.

Sounds like an awesome business if you're getting the money.

A good deal if you can get it, but probably not the ideal outcome if you’re a citizen paying for it or dealing with systems engineered by them.

Not really assuming they are the prime contractor all the value is in the contracts and those are time limited before being re-competed lose all your contracts and your business value goes to zero because there is no IP like a Google or a Microsoft would have that is at the core of the value of the company.

Many people think palantir cracked AI.

While they are just using tensorflow and other things like anyone else.

Additionally, large tech can't apply for defense contracts because of internal ethical concerns of employees. So the naritive remains.

Who thinks that? I've never heard this. People thought Watson was a lot more capable than it actually was just because of the Jeopardy PR stunt. People outside tech were buzzing about IBM and Watson. In my experience, people outside tech barely know Palantir exists.

A lot of retailers buying their stocks and blindly following people like Cathy Woods and random YouTubers.

Not me, but i see some guys trying to spread that naritive.

"Just using tensorflow" is reductive. TF is a design tool, what matters is what you build with it.

Sure.

But having 'cracked AI' means you are the innovator. Not just using the tools ( eg. Compare that with Google & AlphaGo ).

We're not talking about the same things.

>> even in it's heyday Watson was 95% data science consulting firm and 5% actual valuable technology

So, like a lot of other AI and ML companies out there?

> I really think Watson is one of the biggest tech marketing bamboozles of the 21st century.

The weird thing was that to anyone following ML this was detectable bs.

Reply on news.ycombinator.com