Slashdot Mirror
The World's First Graphical AI Interface (fastcodesign.com)
FastCompany reports: Machine learning and artificial intelligence are so difficult to understand, only a few very smart computer scientists know how to build them. But the designers of a new tool have a big ambition: to create the Javascript for AI. The tool, called Cortex, uses a graphical user interface to make it so that building an AI model doesn't require a PhD. The honeycomb-like interface, designed by Mark Rolston of Argodesign, enables developers -- and even designers -- to use premade AI "skills," as Rolston describes them, that can do things like sentiment analysis or natural language processing. They can then drag and drop these skills into an interface that shows the progression of the model. The key? Using a visual layout to organize the system makes it more accessible to non-scientists.
Visual Basic for AI
You are welcome on my lawn.
Give it a graphical interface and hand it over to a bunch of "UX engineers".
Pretty soon it'll have a pastel Fisher-Price interface copied from Chrome.
It's called lego mindstorms and I used it to build a beer pong robot in the college
come on, world's first? There are many GUIs about these topics.
If it doesn't have all the possible skills, would it still be intelligent?
Oh, I see, this is just in case you just want to build a PI (President 'Intelligence') where you'd just need a small subset of skills.
While that's nice and all, neural network programming isn't that hard. Really.
I can put something together in an evening using pre-trained networks, and I'm a huge hack.
Fiddling with hyperparameters is fiddly and time consuming, but it's just bussywork really.
I'm currently pursing a Ph.D. in NLP. Those dirty rotten AI researchers are after my soon-to-be-AI-research-job before I've even graduated! O, wait...
That sounds really great, but can it communicate with Serenity and let them know their show was murdered?
That looks a lot like the tablets on Westworld.... or the other way around?
this is not just about here is tool but only the "very smart" can use it. Such egotistical bullshit and hubris.
I will happily avoid working on a complex system to work on a simpler one simply because there is a lower barrier of entry for it.
It does not matter what you are doing people will elect the path of least resistance just like electricity. If you make your product hard to use, then it will only be used by people required to use it because going without is not an option. The complexity is really NOT the immediate issue. If you make your product easy to use then no matter how complex it will get later you will find a much higher rate of adoption AND understanding of the product as people become more familiar through exposure alone.
After all, would you rather have a car that you could just put a key in and turn on and drive with an automatic transmission, or do you want the one where you have run to the front and crank to start and a manual shifting transmission?
I know this!
#DeleteChrome
... will be revenge for making AI's learn JavaScript.
It's customary to give these things a recognizable, albeit slightly unusual, female name.
One can use UML for modelling.
My experience is that getting the data ready for the "math" part is about 80% of the work, and you have to know something about the models to know how to prep the data. For example, a recent text classification I did with Python and Keras has about 10 lines of code to define, train and test the neural network - but a whole lot more code to extract the data I needed and then beat it into shape for the modeling step.
That said, I'm quite happy with "Python for AI', as it's quick and simple to do things. Please don't make it suck like Javascript :-)
I studied AI at university some 15 years ago. I wrote a program called Polycow. It was an implementation of genetic algorithms with user interactions. You could see how the cows changed as they bred and also how their temperaments changed over the generations. For my own ease of use during the project I wrote a front end for it (all done in Java) that allowed you to set the parameters by choosing icons for aggression, sassyness, kindness and humour. I see no reason why that little project couldn't have other AI modules plugged in to it, maybe you could breed in a chess playing trait for example, or some path finding.
The point: this does not sound like the first of anything.
If it was really Artificial Intelligence you could just say "computer, do so and so" and it would figure out for itself how to do it.
This posting is provided 'AS IS' without warranty of any kind, implied or otherwise.
are 2 different things.
It bears a much closer resemblance to texture processing in 3d modeling apps.There may be other specific domains that have this sort of approach, but web development is certainly not one of them.
XML is like violence. If it doesn't solve the problem, use more.
Machine learning and artificial intelligence are so difficult to understand, only a few very smart computer scientists know how to build them.
Which, amplified by 'fantasy' input from television, movies, fiction, and the media, is probably why apparently the vast majority of people over-estimate the capabilities of so-called 'AI', attributing capabilities to it that it does not and cannot posess, and trust the technology way too much.
The tool, called Cortex, uses a graphical user interface to make it so that building an AI model doesn't require a PhD.
Great. Now we'll have unqualified people, with an almost total lack of understanding, believing they can be 'AI Scientists', creating even more half-assed excuses for AI. Wonderful.
Coming soon to a Best Buy near you: Do-it-yourself Home Nuclear Reactor kits! What could possibly go wrong?
Make their neural networks toolbox usable through simulink as well.
Well, this is total bullshit:
Machine learning and artificial intelligence are so difficult to understand, only a few very smart computer scientists know how to build them.
I'm an untrained self taught javascript coder and I play with neural network scripts all the time. I didn't realize I was so special!
I see pictures with too little resolution to read almost all of their text. I see no links to any demos. I see an UI which is claimed to be awesome but instead draws linkage lines directly on top of the text of other elements. I see an app store where they can rake in a ton of profit. I expect this system to be hyped like crazy and then never be usable.
The author (or the company's media folks) know so little about AI that they mistake the categories of AI for actual things which can be built. You can't build "machine learning" but you can build something which uses machine learning or a framework for doing ML. The author isn't an expert in this field, so their claim of the first GAI is completely untrustworthy. They're also using the term incorrectly. A Graphical AI Interface would be a graphical interface which is used by an AI, not an interface to an AI system. Just like GUIs are graphical interfaces used by a user, not an interface to the user.
There is nothing good in the article. It's all fluf, buzz words, and misdirection with no proof of any of it.
bought to you by msmash: the fountain of drivel.
Slashdot has taken a huge dive. Literally! Look at their web rankings. There are much better science and tech news sites out there.
Given that in my old job we developed one. First well-known one, perhaps.
What these dudes making AI with a PhD are doing instead is a new level of bullshitting with fancy words that impresses people with money and of course legislators who are about as clueless regarding computer technology. They think manipulating a URL to look at the image directory of a server is "hacking".
Machine learning isn't all that complex and it sure isn't even new either. I agree with others here that this is just an ignorant journalism major spouting off buzz words.
If you want to see a really nice GUI designed AI interface? Grab Scratch from MIT and then look at some of the AI experiments that have been done in that programming environment. They aren't necessarily all that fast and certainly some other programming environments would make them work more efficiently, but it isn't even all that new.
Also.... the other shoe dropped when they got into the "app store" business model the developers of this "Cortex" programming environment started to explain what they were doing. It is a scam to separate you from money in your wallet where the author bought into the buzz words to make this seem like a cool thing.
SPSS Clementine does the same thing. It's been around since 2005. The claim to be the first to do this is just BS.