The Alexa Prize
developer.amazon.comAmazon is notorious for this kind of challenge. They offer tiny rewards for huge breakthroughs and then they own any IP submitted. The picking challenge is like this.
"Each Entrant hereby grants Sponsor and its affiliates a non-exclusive, perpetual, irrevocable, worldwide, and royalty-free license to make, have made, use, sell, offer for sale, import, export, license, exploit, promote, reproduce, make available, publicly display, publicly perform, create derivative works of, and otherwise exercise all intellectual property and other rights in and to any concepts, works, inventions, information, designs, programs, or software that Entrant or his or her Entrant Team develop or submit in connection with the Competition or the creation of the Socialbot,"
Don't let your talented friends throw their IP into this pit!
Yup, I keep seeing Amazon trying to run these types of events that show very little respect for the intellectual property generated, creativity employed, and time and effort expended.
They are in full on "suck" mode for their own self interest.
Meanwhile throwing tons of money at another thermostat company.
Yawn.
http://www.businessinsider.com/amazons-alexa-fund-has-invest...
Amazon need to realize that from the outside they appear cheap, arrogant and out of touch with developer mindset.
The Alexa is a good product and worth throwing a lot of money at.
Jeff should be doing whatever it takes to curate and maintain talent and show that customer focus extends as much to the development community as it does to my mother buying windex.
They really need to get serious about developer relations. It's not like they haven't had sufficient warnings, or the opportunity to do something about it.
Well, to be fair, non-exclusive means Amazon gets to use the developed tech in any way. The authors of the IP retain the right to use the tech in any way they see fit too, including making a competing product or applying the tech to novel domains or licensing it to a competitor. Of course if Amazon wanted to apply the IP to a novel domain they could too.
This is a better than most employee deals where the employee has no rights to the developed tech. Even most contract work gives the company exclusive rights as a "work for hire".
It's more like an MIT license between the developer and Amazon rather than a proprietary license just for Amazon.
Wow, that's pretty bad.
I applaud Amazon for the offer. That being said…
> Amazon will award the winning team $500,000.
This is hugely under-paid for a worthy winner. AI startups are sold way more these days; Amazon would benefit tremendously from a novel technique in that domain.
> Additionally, a prize of $1 million will be awarded to the winning team’s university if their socialbot achieves the grand challenge of conversing coherently and engagingly with humans on popular topics for 20 minutes.
This emphasizes further how much those students would be ripped off. Their success is valued at half that of their university, even though they are already paying heavily for that university.
More to the point, a bot that fits this description is a major achievement, beyond Siri and Cortana — which both have a much, much larger value than a million dollars.
I understand that the point is to convince universities to grant their students more time to work on that project, and universities tend not to care about pocket money. However, this ⅓ / ⅔ cut is unbalanced.
>> Amazon will award the winning team $500,000.
>This is hugely under-paid for a worthy winner. AI startups are sold way more these days; Amazon would benefit tremendously from a novel technique in that domain.
They're not paying for a startup with developed tech, established PhD researchers, and a business plan; they're paying students to do research for a year. Assuming a team of 5, $100K/year is very competitive in the academic world unless the students are at the stage of their career where Google recruiters are calling them every day. The likely result of this is something that will have interesting results and maybe a nice demo but will need further refinement in the industry apparatus in order to reach the market.
>> Additionally, a prize of $1 million will be awarded to the winning team’s university if their socialbot achieves the grand challenge of conversing coherently and engagingly with humans on popular topics for 20 minutes.
> This emphasizes further how much those students would be ripped off. Their success is valued at half that of their university, even though they are already paying heavily for that university.
Indeed, this definitely points to the larger imbalance in academic research finance. My university takes a huge proportion of the grant money that faculty win, and reduces aid for graduate students if they independently acquire outside scholarship money.
Divide that $100k/year by the number of entrants and you have the expected payout per student.
I'm fairly sure there are going to be more than three entrants.
>This is hugely under-paid for a worthy winner.
I'm still reading through the competition rules, but it's not clear to me if the winners lose their intellectual property rights. In some past tech competitions I've seen the winners still got to keep their product, the point of the competition was just to get people interested in the domain.
>AI startups are sold way more these days; Amazon would benefit tremendously from a novel technique in that domain.
Are they really though? I've been experimenting for a little bit with neural nets for chatbots. It's always been an interest of mine. I never thought it would have much economic value or interest until now.
> This is hugely under-paid for a worthy winner. AI startups are sold way more these days;
Sure, but AI startups also crash and burn. This seems like a lower barrier to 500k than going to a startup.
> Additionally, a prize of $1 million will be awarded to the winning team’s university if their socialbot achieves the grand challenge of conversing coherently and engagingly with humans on popular topics for 20 minutes.
This does bother me too, but I can see how it incentivizes universities to up their game.
Universities already have incentive to up their game in the form of being an university. The potential to get a million dollars if they maybe win some contest is a complete abstraction to the people who manage the establishments finances.
But, does Amazon pay the $500k even if the winner doesn't actually beat current state of the art? Seems like you just have to beat out the other competitors.
In that case found a company and exit by selling to AMZN for $FAIR_MARKET_VALUE.> This is hugely under-paid for a worthy winner. AI > startups are sold way more these days; Amazon would > benefit tremendously from a novel technique > in that domain.that's not a fair comparison - you would need to "found" many companies (comprising of top grade university boffins)..
This really is a cheap win... for Amazon.
It's Amazon, what do you expect?
The intellectual property section towards the end of the rules[1] seems noteworthy (though not unexpected).
>Each Entrant hereby grants Sponsor and its affiliates a non-exclusive, perpetual, irrevocable, worldwide, and royalty-free license to make, have made, use, sell, offer for sale, import, export, license, exploit, promote, reproduce, make available, publicly display, publicly perform, create derivative works of, and otherwise exercise all intellectual property and other rights in and to any concepts, works, inventions, information, designs, programs, or software that Entrant or his or her Entrant Team develop or submit in connection with the Competition or the creation of the Socialbot, including any of the foregoing included or described in any Technical Article or other materials provided to Sponsor.
At least it's non-exclusive so they can take their tech elsewhere too.
> create derivative works of
Interesting, seems like they could publish your source under whatever license they chose. It might be in their interest to use a BSD/GPL/MIT license to continue community contribution. Though it might hurt your ability to turn that software into a Conversational-AI-aaS business.
So if you have something capable of winning this prize, how much more would it be worth to keep it?
Whatever OpenAI might offer, which I'd argue is a better steward of the resulting knowledge than Amazon.
If OpenAI would release an Alexa hardware competitor, I'd pay whatever it cost in a heartbeat, as I'd know it was being used to increase the data corpus of a non-profit versus Amazon.
Of course, it depends. How well can you market/sell it? How much is winning this prize worth to you in future job earnings?
I'm curious why only students can apply? Why not allow individual developers or companies as well?
Because Amazon wants good PR and university pipelines for promising future employees, not subsidized contractors.
Churn and Burn
Think about what Amazon's turnover must be compared to other companies'.
Apparently, the Median Amazon tenure was only 0.1 years shorter than Google's in 2013: http://www.businessinsider.com/companies-ranked-by-turnover-...
Although interesting that Google's turnover rate is close to Amazon's, one year turnover is horrible. Other large tech companies (like Microsoft and Apple) have much lower turnover rates.
Tenure is not turn-over. If you double the size of your company every year, but nobody quits, then your average tenure is <1 year. Nobody would be able to accuse you of having high turn-over, though.
On the other hand, if you grow your company by 25% year-over-year, and your entire staff quits after 18 months, your average tenure will be >1 year.
Amazon thrives on new grads. Get a bunch, work them to the brink for 2 years, only the "strong" survive to become senior, repeat forever.
Is it just me, or does 1 month seem like a really short time frame for the initial invocation of this contest? Why couldn't they set the deadline farther back?
If you possess this technology, you can sell yourself to either Amazon, Google or Microsoft for potentially hundreds of millions of dollars.
2.5 million dollars for such capability is by far the most abusive prize ever.
If you have seen "Pirates of Silicon Valley", you would be the equivalent of the guy that sold DOS to Microsoft for $50,000.
I think this program is great and similarly reminds me of Google Summer of Code. The biggest deficiency I see with new engineers is lack of real experience. GSOC, this Amazon program, these give students the chance to build a real thing that does something. Too many students come out of school with their theoretical assignments completed, and basic hello world programs.
I hope Amazon receives great submissions from this!
Isn't the Summer of Code for working on an existing open source project?
This competition is about producing a SocialBot that Amazon will exploit.
The open source projects are not exploiting in the same way? I am not sure I see your point. Amazon is offering prize awards for projects. I think more companies should start offering these types of opportunities for students to garner industry experience.
My point is that the two programs are not so similar.
One is about bettering community projects, the other is explicitly about creating projects for Amazon to commercially exploit. I'd say it is similar to other corporate hackathons.
By exploited you mean release to the public domain. Big difference.
I didn't know the Turing Test was renamed to Alexa Prize
If the Turing Test stipulated that the human were a politician, it would have been beaten many years ago.
> You know, that's a very interesting point. I tend to agree with what you're saying, and I'd go one step further and say that we as a nation ought to be doing more about that.
Ad nauseam. You could call it E-lies-a.
/snark
I beleive this is a critical issue and one we should be having a broad national debate about.
From their FAQ:
>Will this competition be judged like a Turing Test?
>No. The goal of the Alexa Prize is to create socialbots that engage in interesting, human-like conversations, not to make them indistinguishable from a human when compared side-by-side. While the socialbots built for the Alexa Prize will be human-like in some respects, they will be very different in others, and could easily reveal themselves in a Turing Test. For example, socialbots may have ready access to much more information than a human. Asking the socialbots to act human could diminish the customer experience and hinder the efforts of the participants to build the best socialbot to further conversational AI.
Ooh, so they expect the bots to fail a Turing Test because they're superhuman? That makes it easier. /s
No they didn't say that. A socialbot should fail a Turing test because no human could Google an answer in 1 second or know the exact temperature, etc.
A human plus a googlebot could do it though! Have the bot scan the incoming messages, provide possible facts related instantly. Then the human writes the response. The Turing test would say Robot but it would not be entirely right.
The point of the Turing test is for an AI to pretend to be human so well, humans can't tell if it's human or not. The bots in the Alexa Prize do not have to pretend to be human, and can be very different from humans. The point is just to make bots that are interesting or useful to talk to.
It'd be cool if someone built a bot that would help you practice another language. I spent years learning Spanish but don't have a partner to speak to in the language.
Everyone is complaining that the prize is very small relative to the value of a successful dialog startup.
But I don't see what's bad about this. Amazon is open about how small the prize and presumably that will mean that anyone who has a shot at making a dialog startup won't participate.
So Amazon will probably get a lot of entries from groups that are sort of borderline, which isn't necessarily the worst thing in the world.
Yes it is quite laudable how transparent they are about being cheap SOB's
Limited to students only?
I guess they wont be sharing the dataset of actual alexa searches so far.
Geez, I hope not. Remember the disaster when AOL Research thought releasing search history was a good idea?
Come on, $500K is a pittance to the winner which could create a compelling new use case for Alexa. Amzn, you can do better and dont be penny wise, pound foolish
They keep using the word "socialbot". I've read through the FAQ and I still only have a vague idea of what they have in mind.