This is great of course, but like with any programming language, it will only succeed once there is an ecosystem around it of libraries/packages/api's/frameworks and a critical mass of information about it on websites like Stackoverflow. So often once a language is released, it might still take a decade until the ecosystem is developed and mature enough.
Language literally comes out yesterday... Here's a new job offer tomorrow.... *NEW JOB OPENING... CANDIDATE MUST HAVE:* - 8+ years Python experience as senior developer with django, flask, pyramid, AI, and ML - 5+ years JavaScript experience - 4+ years Java experience with Spring framework - 10+ years Certified Linux Administrator - 10+ years Gen experience for ML and AI - Pay: $45,000 US dollars / year
In their own words, it is for non technical types to PROTOTYPE an AI model. This prototyping can already be done in Python in maybe 4 to 8 lines of code. The real trick is get an AI to production quality. For "hotdog-not hotdog' at 82% accuracy no problem .The real trick is robust models which can work in real world situations without problems. For this you are going to need Maths and lots of it.
Adoption and the accuracy of the models is going to determine if this language succeeds or not. The reason why math is so important in ML is because accuracy of the models. If the models are not accurate enough it will be another niche language. We will see.
Ok sorry to say this but math is still important for AI. Part of the reason why PhDs are getting hired for machine learning work is because a lot of machine learning work requires the same skillsets as academic researchers, like recreating models and experiments from academic papers (sounds like a lot like the skills needed in a researcher). Once the models are built, then the engineering part kicks in which is to retrain and deploy the model to be used for a different application. Whether you're retraining a model (which means you need to know how it works in the first place) or you're developing the model, you still need to understand math, especially calculus, linear algebra, and probability in order to understand these machine learning papers.
Thanks a lot! Any chance you could consider making a video in regards to blockchain? Maybe Solidity even! It is out of a lot of peoples comfort zone and it seems that it is coming rather sooner than later and being prepared is preferable :) Comfort zone is bad, it stops innovation and creates the box which is hard to leave.
And what is the point of this language anyways? You can already build a model in like 5 lines of code if you use Python, tensorflow, and keras. The one thing about tensorflow that I've found that makes it incredibly powerful for what I do is that it allows me to build, train, and test a model in python and then deploy it in any language (I build my models in python and then deploy it in C++). Unless Tensorflow decides to rebuild everything around this language, I personally don't see this language becoming a mainstream language for machine learning.
@@manishsinghbisht3010 good question, unfortunately I wouldn't know because I've never benchmarked the performance. Although if I were to take a guess, I would imagine there would be some performance cost if you were to use tensorflow's savedmodel format to save the model in python and then load it back into C++, but the slight performance cost probably does not justify the extra time and money it will take to build the model from scratch in C++.
Microsoft released ML.Net if you want to use ML in say C# code and Infer.Net has been around for years . Other than that Python has had pretty popular ML and Data Science libraries for years. I never heard of Gen I looked it up it looks interesting looks like it has some relation to Julia which is a fast data science language which I heard about at a R programming user group once years ago.
...relax people, Gen isn't a language. It's merely a new ML package for Julia, one of many such as FLUX, Knet, etc. Julia, which I predict will overtake Python as the lingua-franca for AI & ML, is an incredible technical achievement in such a brief period of development time. Kudos to the Julia team at MIT!
@@cjb1373 That's not a sequitur from my comment, but having said that - personally I'm very skeptical of the entire AI field as a panacea for the future of computing. I've gone through lisp, Prolog, Expert Systems, Fuzzy Logic, and so-on - they're all tools of varying efficacy, but I think modern AI thinking revolves around too many movie scenarios. WOPR anyone?
This is great of course, but like with any programming language, it will only succeed once there is an ecosystem around it of libraries/packages/api's/frameworks and a critical mass of information about it on websites like Stackoverflow.
So often once a language is released, it might still take a decade until the ecosystem is developed and mature enough.
probcomp.github.io/Gen/dev/ Documentation here, if anyone wonders.
Language literally comes out yesterday... Here's a new job offer tomorrow....
*NEW JOB OPENING... CANDIDATE MUST HAVE:*
- 8+ years Python experience as senior developer with django, flask, pyramid, AI, and ML
- 5+ years JavaScript experience
- 4+ years Java experience with Spring framework
- 10+ years Certified Linux Administrator
- 10+ years Gen experience for ML and AI
- Pay: $45,000 US dollars / year
lol xD
In their own words, it is for non technical types to PROTOTYPE an AI model.
This prototyping can already be done in Python in maybe 4 to 8 lines of code.
The real trick is get an AI to production quality. For "hotdog-not hotdog' at 82% accuracy no problem .The real trick is robust models which can work in real world situations without problems. For this you are going to need Maths and lots of it.
Adoption and the accuracy of the models is going to determine if this language succeeds or not. The reason why math is so important in ML is because accuracy of the models. If the models are not accurate enough it will be another niche language. We will see.
Always delivering values , thank sir.
Thanks so much for your videos! So glad to watch such content quality you present!
New languages just never ends
True.
How is this different than what we have in R and Python? What are they missing?
Ok sorry to say this but math is still important for AI. Part of the reason why PhDs are getting hired for machine learning work is because a lot of machine learning work requires the same skillsets as academic researchers, like recreating models and experiments from academic papers (sounds like a lot like the skills needed in a researcher). Once the models are built, then the engineering part kicks in which is to retrain and deploy the model to be used for a different application. Whether you're retraining a model (which means you need to know how it works in the first place) or you're developing the model, you still need to understand math, especially calculus, linear algebra, and probability in order to understand these machine learning papers.
What exactly do I need from linear algebra? Currently enrolle in that course
i wonder whats gonna be tensor flows use now
This is where being over 300 years old has an advantage.
I want to start learning this asap.
"WIth Programmable inference" not interfaces....
Thanks a lot! Any chance you could consider making a video in regards to blockchain? Maybe Solidity even! It is out of a lot of peoples comfort zone and it seems that it is coming rather sooner than later and being prepared is preferable :)
Comfort zone is bad, it stops innovation and creates the box which is hard to leave.
He's already made a video about blockchain and bitcoin a while back.
Where's this AI programming language now? I've never heard of it, neither in 2019, nor since then.
With programmable inference and not interfaces ^^ Math jargon here stef =)
I just searched the comments to see if this one existed or if I would have to create it. xD
@@ThePublicTutorials Hahaha. Took care of it thanks :)
And what is the point of this language anyways? You can already build a model in like 5 lines of code if you use Python, tensorflow, and keras. The one thing about tensorflow that I've found that makes it incredibly powerful for what I do is that it allows me to build, train, and test a model in python and then deploy it in any language (I build my models in python and then deploy it in C++). Unless Tensorflow decides to rebuild everything around this language, I personally don't see this language becoming a mainstream language for machine learning.
@@manishsinghbisht3010 good question, unfortunately I wouldn't know because I've never benchmarked the performance. Although if I were to take a guess, I would imagine there would be some performance cost if you were to use tensorflow's savedmodel format to save the model in python and then load it back into C++, but the slight performance cost probably does not justify the extra time and money it will take to build the model from scratch in C++.
Very interesting...thank-you for the awareness of MIT's Gen!!
Have to stay in the loop.
The life of a coder / programmer / computer scientist, @@StefanMischook!! =)
Where can I learn "Gen", the AI programming language?
It's good to hear that AI is easier to get into. Maybe we'll try getting into that
Microsoft released ML.Net if you want to use ML in say C# code and Infer.Net has been around for years . Other than that Python has had pretty popular ML and Data Science libraries for years. I never heard of Gen I looked it up it looks interesting looks like it has some relation to Julia which is a fast data science language which I heard about at a R programming user group once years ago.
Great!
...relax people, Gen isn't a language. It's merely a new ML package for Julia, one of many such as FLUX, Knet, etc. Julia, which I predict will overtake Python as the lingua-franca for AI & ML, is an incredible technical achievement in such a brief period of development time. Kudos to the Julia team at MIT!
Ha! Who needs math? We've always had calculators anyway.
Exactly!
Total scam python AI libraries are easy to use and you get everything python has
In my opinion, if you want to make money and work in the industry - ignore this language.
Does this mean it will be easier to become a machine learning, AI engineer without a degree?
@@cjb1373 That's not a sequitur from my comment, but having said that - personally I'm very skeptical of the entire AI field as a panacea for the future of computing. I've gone through lisp, Prolog, Expert Systems, Fuzzy Logic, and so-on - they're all tools of varying efficacy, but I think modern AI thinking revolves around too many movie scenarios. WOPR anyone?
@@Jollyprez ...I agree. Let's see where it goes from here.
For all PHP haters: LARAVEL!!!!!!!!!!!! (PHP hater gets roasted)
Math is hard ! please just git rid of that in every aspect of computer users/programmers :D