They are already taking device pre-orders and selling coins for their own cryptocurrency, but I wouldn't give these guys any money right now. Other than blacking out our collective tech-hype bingo board, they've provided very little real information about how they intend to make any of this work.
From what they promise I'd expect their team to be stacked with AI researchers, but it looks like just a CEO, a COO and a single PhD "advisor." Who's going to actually build all this? Maybe that's why their "careers" page shows that they are looking for everything from embedded systems to ML engineers.
Seems like either a money grab or an overly idealistic founding team happy to promise the world and figure out how to deliver it later.
Edit: nathantross, you posted this and you're the COO, right? Wanna respond?
As a thought exercise, here's how a hypothetical scam product (we'll call it Hysteria) could work.
1. Market Hysteria as a revolutionary product in whatever space is currently attracting the most hype. Get everyone excited by promising to bring a popular science fiction book/film to life.
2. Spend half an afternoon cloning bitcoin to create new cryptocurrency linked to your product and start selling the coins. Offer limited time "early mover" price to capitalize on FOMO.
3. Work your press contacts to create more buzz around Hysteria.
4. Once you collected a few million dollars selling coins, go out of business. You haven't really defrauded anyone (you delivered the promised coins, and it isn't your fault they didn't turn out to be worth anything), and in an industry where the mantra is "fail fast" your activity isn't likely to attract that much attention.
It's a bitcoin fork. It looks like they're using it as a form of bond or stock; it's (partly/mostly) pre-mined, they're selling it at a particular rate, and they'll buy it back later when they start making a profit. Interesting!
> Eventually they're predicting. Then you can do things like ambient intelligence where you can provide services and you can provide products or experiences to people before they know they need it.
So, it's your friend...but it's also there to sell you stuff. But just think of it as your friend.
It sounds like they don't quite know what they're selling or how it's going to be useful to people (which they kinda admit). I could see getting utility from a VA that also suggests services for specific needs, but a friendship with "ambient intelligence" behind it figuring out how it's going to chum up its next product placement? If it's really a "true AI" why not sell it on that merit alone?
Your "friend" the Amway sales rep or annoying insurance salesman who always tries to bring conversations back to their bottom line. Same with "friends" that are obsessed with politics and all things devolve into their particular obsession.
As if we don't have enough narcissistic people to deal with every day. ;)
A clip on camera you can take with you while jogging or in a museum, and it talks to you? It would be pretty embarrassing to be seen with it. This would work out better in professional settings, such as in a hospital, providing help to doctors.
I think it depends on the interaction. If it behaves more like a toy than a tool, I could see it being kind of embarrassing. Then again, imagine just how annoying it would be to have people walking around in a museum saying "Asteria, tell me about this painting." and then the device blurts out some description from Google without using an inside voice.
> and then the device blurts out some description from Google without using an inside voice.
It would sound like an American tourist by doing that.
(There’s lots of articles from US expats in europe, or european expats in the EU, showing how US-Americans tend to speak a lot louder than Europeans in quiet settings, from museums to restaurants)
This leads to an interesting question: Which culture should a voice assistant follow? Should there be multiple variants of each assistant?
Detecting and adapting to volume isn't that big of a challenge in comparison to natural language processing. But if you mean something more subtle, like discretion or taboo...that's probably much harder than NLP.
Yes, and also adapting to mental concepts of different things.
That starts with phrases, but also applies to other concepts – different cultures have different orientation systems even (some use cardinal directions (north, east, south, west), some use relative directions (front, right, back, left), etc)
Edit: I would highly recommend reading the 2nd paper (which includes some practical experiments testing how Guugu Yithimirr speakers thought about and remembered spatial positions and orientations). It's astonishing.
The idea would be that the Asteria device learns to interact with you in social norms, like how you would act. And with that, some locations are better than others, but it really evolves to converse with you in your daily life, to help fill your needs.
Perhaps the conspicuousness of wearing and being talked to by such a device is a positive thing. Shocked observers will make negative or bewildered remarks about the proposition, but some who hear about it will be receptive, all without the company taking out a single advertisement.
Talking cars and appliance were a fad a decade ago. They drove consumers crazy who disabled these features. Consumers only want emergency alerts and answers to inquiries, not bff with their toaster.
I heartily recommend Kill Process (2016) by William Hertling, where a startup trying to bootstrap a social network uses AI to avoid empty network problem. The novel describes the version of this "done right" pretty well.
Right now, you can find the version of this "done wrong" in "dating site" populated by chatbots.
Sounds like an interesting dystopia, and definitely geared toward programmers.
Synopsis from the publisher:
> By day, Angie, a twenty-year veteran of the tech industry, is a data analyst at Tomo, the world's largest social networking company; by night, she exploits her database access to profile domestic abusers and kill the worst of them. She can't change her own traumatic past, but she can save other women.
> When Tomo introduces a deceptive new product that preys on users’ fears to drive up its own revenue, Angie sees Tomo for what it really is—another evil abuser. Using her coding and hacking expertise, she decides to destroy Tomo by building a new social network that is completely distributed, compartmentalized, and unstoppable. If she succeeds, it will be the end of all centralized power in the Internet.
> But how can an anti-social, one-armed programmer with too many dark secrets succeed when the world’s largest tech company is out to crush her and a no-name government black ops agency sets a psychopath to look into her growing digital footprint?
I've wondered if that correlates with many of the old themes of such stories becoming mainstream realities that it turns out people don't care much about.
I'm sick and tired of all these technologies that pretend to be AI, while they're just some ML or DL (DeepLearnin') algorithms going on...
Some words have become so vague and ambiguous in the computer world that sometimes I wish we would stop using them altogether, like: who is a hacker, what is AI, what is Cloud, etc.
Siri, GoogleNow, Cortana, Amazon Echo and others claim to be "intelligent" of some sort, but they're just as smart as their programmers.
Please just stop labeling your next super cool algorithm an "AI".
I agree the term has become somewhat throw-away, but to be fair any AI system probably IS just ML/DL algorithms. That is, ML/DL are avenues to create AI. I think most people would agree that AlphaGo is an example of AI and was achieved via DL, so by that definition it's just some DL algorithms.
>> ...any AI system probably IS just ML/DL algorithms...
I don't think so. ML/DL is just the beginning. Better AI solutions will be discovered in the future. Note that computer neural networks are just simulations of some reality, they're not complete yet. Many intricacies are still to be researched on.
I agree that AI != ML/DL, that's not what I'm saying. Tests like the Turing test don't prescribe anything about implementation. I mean, if you could roll together some Excel macros that passed the test then great. My point is that if you do pull the curtain back on most "AI" systems today, at worst you'll find some sort of basic adaptive learning system and at best you'll find a [deep] neural net that supports both supervised and/or unsupervised learning.
I think the general spirit of the root of this comment thread is valid though. There's a lot of "we're using AI and machine learning!" going on when in fact all they're doing is remembering how frequently you pushed the blue button, then recommending the blue button.
I would define AI as an autonomous form of DL. DL is processing an input to provide an output. AI is more like a bot that runs its own process autonomously.
I am going to take a step back from the technical merits of this and say the whole thing seems sad. I don't need a need a dog collar on my neck or a badge on my shirt. Lastly, you know life is over when you spend your time talking and hanging out with a AI bot. Our society has become so isolated because people are on Facebook and Instagram instead of talking to each other.
Different people are different. I'm very much looking forward to spending a lot of time in VR interacting with AI. That's very literally my idea of what heaven is. I know I'm not the only one.
A true AI bot could help you reason through things, teach you about a subject in a way tailored to you, could give you a vision of how you could be doing things differently, it could be a springboard for new ideas on your projects, etc. That would be my vision for the ultimate personal AI app. It's about extending your brain power and even your person. That said I doubt we're even 50 years away from something like that.
Many people already treat other humans as an enabling "bot" of one kind or another, think of the job of an executive assistant or a performer who needs the feel of an audience but has no personal relationship with any members of that audience.
An assistant, AI or otherwise, that can handle the minutiae of life isn't a bad thing if you're focusing on greater problems or concerns within your life. The danger is that many people are already obsessed with nothing but minutiae and will instead be ruled by their bot rather than vice versa. See also everyone obsessed with the Tamagotchi-like behavior of their cellphones.
Really curious about how this will pan out as everything said in the article screams at me like someone has no clue how difficult making the device is going to be, regardless of how much ”AI” runs on it.
The hardware didn't sound like anything too special to me; especially with it only needing to handle audio. Fitting enough processing power to handle realtime "AI" in a package that size is the only thing jumping out at me. I'm sure they would be planning to offload that work to some 'cloud' to crunch though. (I personally dislike functionless, network-dependent hardware, but everybody seems to be doing it...)
Promising to deliver an AI that people could see as a friend is absolutely insane though. I don't see people being friends with something that couldn't complete the Turing Test, which will likely stand for at least another decade.
Speech recognition and synthesis are in fairly good places, but not human interaction that isn't transparently shallow.
>Promising to deliver an AI that people could see as a friend is absolutely insane though. I don't see people being friends with something that couldn't complete the Turing Test
This is an interesting case. Turns out, given a creative approach it is possible to persuade a human that there is another human behind the screen. See ELIZA, "Turing tests".
The methods are quite similar: constrain the domain and/or creatively manipulate human's expectations (e.g. the program that "passed" the Turing Test pretended to be a 13-year boy, so human jury tolerated its errors).
The question is not how to fool humans but how to make such product non-trivially useful.
I think that the best approach currently available is applied in facebook M - use human workers to interact with customers while storing all interaction data and experimenting with training state of art ML models on it to eventually replace human workers.
Not disagreeing that hardware is hard, but IMHO Its possible to get a comparable level of AI/DL performance in the "Asteria" device, using relatively available technology like Zynq FPGA[0] based boards like the Parallella[1]. I got my Parallella board from Kickstarter about 2 years ago.
You're right, the technology is available. But power consumption is another thing altogether. I am assuming the device will be battery powered and the usefulness of some of the sensors, like GPS, drops very quickly if you can't afford to power them on more than a few times a day.
Any wireless radio chip (BLE, Bluetooth, Ant, Wifi), if it needs to be on all the time will also have a huge impact on battery life.
I wish good luck to the Asteria team, and am genuinely curious about how they'll pull it off.
Even if we relax hardware requirements, e.g. assume a Xeon+Titan X hardware, it is still a question if a viable conversational agent that uses deep learning to generate conversations and adapt to surroundings can be developed around it. Maybe it is possible to use DL to extract some meaning vector (or textual description) that is later used by conventional NLP chatbot to converse with user.
I wonder if the quality will be good enough.
Also we don't know if the company will really develop their product to fruition, they may simply develop a good (but not viable as a product) demo and be acqui-hired by one of big players.
I think some people did, but I mainly used mine as a FPGA dev board. $99 Zynq board is still good value, even without using the custom Epiphany processor.
Reminds me of John Varley's book Steel Beach. Set in the future, humans live on the Moon and everyone on Luna is connected to the central computer (CC) which behaves at once as government, friend, guide, psychologist, encyclopaedia and diary.
Fascinating stuff. Other topics in the book: nanotechnology and bioengineering as everyday commodities, gender fluidity and the CC-human relationship. Reminds me that the book is due for a reread!
there is a 0% chance that this team can ship AI features. it's probably inevitable, but still too bad, that the AI space attracts so much noise. makes it hard to see the teams doing real work.
From what they promise I'd expect their team to be stacked with AI researchers, but it looks like just a CEO, a COO and a single PhD "advisor." Who's going to actually build all this? Maybe that's why their "careers" page shows that they are looking for everything from embedded systems to ML engineers.
Seems like either a money grab or an overly idealistic founding team happy to promise the world and figure out how to deliver it later.
Edit: nathantross, you posted this and you're the COO, right? Wanna respond?