Rocket AI: 2016’s Most Notorious AI Launch and the Problem with AI Hype | Hacker News Reader

Rocket AI: 2016’s Most Notorious AI Launch and the Problem with AI Hype

medium.com

152 pointsjjwiseman48 comments

Highly respected AI researchers who were in on the joke publicly praised it. VCs are not idiots for showing interest given the circumstances. If they aren't going to judge AI startups by the opinion of world-class experts, how exactly are they expected to judge them?

Maybe the fact that RocketAI didn't have a product, wasn't solving any problem, or was touting a bogus algorithm with absolutely zero details should've set off some flags. They didn't base their interest off of any value they were creating, it was just unsubstantiated hype.

And the VCs didn't invest, they reached out for a meeting.

If there were any term sheets offered to RocketAI, that would be different. It's completely unsurprising that VCs reached out to a company touted by top AI researchers, wanting a meeting.

The filter for relevance for investment that has overwhelmingly survived in the market is social proof ("warm intros") for a reason: it's substantially more expensive and harder to fake third party signals (including their own quality) than your own.

Why did Google beat all the search directories? The latter used keywords on pages to figure out relevance, which was easily gamed by sites putting up thousands of keywords. Google weighed (still weighs) both the number of links and the quality of each referrer. The SEO vs search relevance war continues to this day, but PageRank gives Google a long term advantage.

The attractiveness of an investment has many variables with the two most important team quality and traction. Many VCs are on record saying they'll fund a pre-product company if the team is worth it.

If a number of high quality referrers are recommending the team, then the team quality variable might override the others. The hype's substance here was the referrer quality.

What exactly is the advantage given by PageRank today?

I agree. I see it as an attempt to somewhat deflate the emerging bubble. Most AI researchers are aware of the effects of the previous 'AI Winters' and probably want to smoothen the hype cycle in order to secure continuous funding in the future.

> didn't have a product, wasn't solving any problem

Since when has that stopped startups? /s

Would a research paper constitute enough proof? Or some "results"? Comparison of their technology to some other state of the art stuff? In other words, what is the proof that VCs (and other AI newbies) should believe?

EDIT: note that VCs just contacted them. That is far from signing a deal.

Yes. I think you're answering your own questions?

My point is that everything can be faked. In the end you have to decide who you trust to pass the judgement, unless you are capable enough to make that call yourself (and that option is just not realistic).

Agreed. It's a bit naive for a bunch of AI researchers who know nothing about VC to be judging a few VCs for showing interest.

First, which VCs? There is a very wide range. And second, which people at those VCs? A lot of associates will reach out speculatively whenever some company or project makes headlines, but there's a ton more diligence and vetting that would go into getting a partner meeting let alone all the way to an investment decision.

For me it sounds like the researchers are saying: we are so cool now we can tease you all we want because we can have 10 investors just by snapping our fingers. Sounds a little like a childish revenge after years of crawling for funding. I'm glad they like to raise awareness about hype but I would choose other way

> Agreed. It's a bit naive for a bunch of AI researchers who know nothing about VC to be judging a few VCs for showing interest.

I see no reason the VCs shouldn't keep some technical experts on staff to make these kinds of evaluations.

Who's to say they didn't?

Perhaps the VCs who reached out were reaching out so that their technical experts could learn about the product. If you can't find information about a startup anywhere, asking them is a great way to find out. Yet the implied attitude of the article and some comments here is that when VCs are missing a critical piece of information, they should just give up, rather than reaching out to gather that critical information. Am I missing something? Reaching out to a startup to gather more information sounds like typical VC hustle.

I think as people who know things about AI that other people don't, we need to be careful about using our knowledge to trick others and then feel smugly arrogant about their ignorance. My opinion, at least.

That would be difficult to do. If someone wants to stay an expert in the field one has to practice it, which rules out 9/5 job at VC headquarters. But most important, it is probably cheaper and more effective for VCs to find an expert when needed, given that they probably need many experts from many different fields.

I can't fault the VCs for checking to see if there's fire when they see smoke. It's part of their job.

I do think it's funny that they spent around $600 bucks and got way more attention than they would have if they bought a NIPS sponsorship for 10k.

It's not a new observation that organic marketing, which can be very cheap, handsomely beats paid advertising.

If you're building a new sneaker brand and you spend $1000 leveraging your contacts to get Drake wearing your shoes at a concert, it'll beat an orders-of-magnitude more expensive advertising campaign.

And to be clear, "showing interest" is all they apparently did. They didn't expose themselves to any actual risk. I wouldn't be at all surprised if they sent off those emails to get a foot in the door, expecting it to not work out, and then started looking around and realized it was a joke. If they'd waited to do their research first, they risked being beaten to the punch.

disappointing seeing those folks getting into this joke. author obnoxious to say the least. More a thing to say "I am over the hype, look how cool I am" than criticize and point out lack of understanding from lay people or tendency to quick adopt machine learning technologies for the sake of this term's cutting-edge idea that it carries.

Maybe by having done some ML themselves. It's really not that hard to get up to speed, maybe a year or so to be conversant and know what is bullshit and what isn't.

I would argue that "a couple people I follow on Twitter said it was cool" is not good criteria for investment. Then again, the post didn't mention any information about how serious the reach-outs were. I'm guessing they may have been introductory emails rather than the "HERE TAKE MY MONEY" messages the author seems to imply.

In other words, some people just discovered the children's joke that you can lie to someone and then make fun of them for being "gullible", destroying your credibility in the process.

Using their brains?

> A machine learning researcher, a crypto-currency expert, and an Erlang programmer walk into a bar. Facebook buys the bar for $27 billion.

https://twitter.com/ml_hipster/status/438418306769244160

I'm about as far away from AI and a developer can be. I found this article to be a fun read. When there is hype, there is really hype.

This should be a lesson to journalist, fan boys/girls and VC. Pay attention to reality.

And here I would have thought that Tay was 2016's most notorious AI launch...

(RIP 3/23/2016 nevar forget, #taken3soon)

Tay wasn't known for how it acted once launched though, it was notorious for what it was vulnerable to/ what it learned after launch.

Tay was only online for a grand total of about 16 hours. I'm sure she was a very sweet, innocent chatbot for the first five minutes, though.

http://i.imgur.com/QYvM2kp.jpg

"1998 – The untested AI is activated for the first time as one of the planned activities on Aperture’s first annual bring-your-daughter-to-work day. In many ways, the initial test goes well: Within one picosecond of being switched on, GLaDOS becomes self-aware. The "going well" phase lasts for two more picoseconds, at which point GLaDOS takes control of the facility, locks everyone inside, and begins a permanent cycle of testing."

I am sure AI is over hyped, but I think this is more of a demonstration of how if you provide people with booze and a venue you can get them to attend anything. A lot of the social media response reads as tongue-in-cheek.

For VCs who focus on AI, I think it would actually be irresponsible if they didn't show interest in a start up that was garnering so much praise from AI experts. Its not as if term sheets were sent.

It also wouldn't be irresponsible to send term sheets - the fact that the company doesn't exist will come out in due diligence pretty quickly.

While I appreciate the trolling, it just feels like people in general are abusing the most important idea in the history of humanity.

Exploring the possible early components (DRN, GAN) that could lead cumulatively to AGI is not a joke. It's going to be more powerful than nuclear or bio weapons. Are those funny?

I used a combination of networks to turn news articles into memes.

I guess I should apologize for not respecting how news2meme could be a stepping stone to AGI, and promise to be more austere in the future.

You don't even want to know what I'm doing with Reinforcement learning and flappy bird.

No those are great uses. Creativity in satire is a necessary trait. My problem is with people who say and then reinforce the idea that AI is just hope it's disrespectful to people who are really working on it.

Yes, we must create safe spaces for these people to ply their trade.

You can't just dangle those in front of us and not post links. I demand satisfaction, sir and/or madam.

(I've seen people do flappy bird with reinforcement learning before, though)

The world isn't ready for a public release of news2meme...aka it doesn't work all that great.

Actually, it's a great idea. I often observe how some news are brought to me via a meme complaining about it on 9gag. 9Gag is really opinion manipulation, because it's a relaxing joke site with opinionated views and without enough room for fact checking — posting counter-news on such a venue could play a major role in future elections. As an example, Clinton hasn't managed her 9gag footprint properly, to say the least (she was vindicated).

Yes, I get that news2meme is supposed to be a joke.

You've been had. The phrase "neural net" has helped many people get tenure, but it's not a model for the human brain.

Regardless, neural nets could eventually lead to AGI.

You could say the same thing about computers. Look, I think neural nets have made significant progress in image, video, and audio processing, but until their practical results are understood at a theoretical level, I'm not convinced that we're any closer to AGI than we were 5 years ago, beyond the idea that "with enough compute power, an intelligence will emerge."

I was just commenting on how you weren't really backing up your statement that parent had 'been had'. He thinks neural networks could eventually lead to AGI. You said he was wrong because neural networks aren't really a good model of the brain. My point is we don't know that something has to be a good model of the brain to lead us to AGI.

There's no reason to believe neural nets in 2016 will bridge the gap more so than folks saying this about nets in 2000 vs in 1990 vs 1980. The introduction of GPUs, more data, and variations in the models have lead to real improvement in specific tasks, but even those get stuck again. The leap to AGI is unclear if that's what's meant. You may be referring to synthesis techniques... but same thing. (They are cool though!)

Again, I'm not making the claim that neural nets will lead to AGI. I was only commenting on the claim that they can't or won't just because they don't accurately model the human brain.

Propulsion could eventually lead to near-FTL human travel.

I think Dr. Strangelove proved doomsday devices are funny.

I'm reading this thinking, "Wait, why is Anders introducing this algorithm... Oh."

If you like Borat, RocketAI and this author are for you.

But for me, it's just Jumping The Snark.

Reply on news.ycombinator.com