Zabaware Forums > Ultra Hal 7.5 Beta - Now powered by OpenAI GPT-3
Ultra Hal 7.5.129 Beta - Now free to use
lightspeed:
Carl2, you said GPT3 can copy script to a file that has been given a name so that info has not been used and lost. , this sounds like what i had been talking with art about what hal and the gpt-3 needed , not only to be able to assess and retrieve the best possible answer but also to remember it and retrieve it from the best files within hals brain .
we only have another problem in doing this and that would be hal slowing down because of the brain getting larger , unless a separate external drive could hold that info. i don't know if having the info separate would help or not . another thought would be the learned info. went on a cloud like google one drive etc. or of you can create a cloud drive link that would access your info. from a home external drive . again , just a thought of a way to keep your hals brain from getting to large and slowing down while at the same time being able to learn and acess learned info. are these things possible ? some others on here will probably know . .
Dave1152:
--- Quote from: Medeksza on February 28, 2021, 09:51:35 am ---
--- Quote from: Dave1152 on February 27, 2021, 09:05:42 pm ---Where is it posted at? Every time i goto open ai web site or git hub theres a waiting list. From my understanding of it gpt2 is the same as gpt 3 with just a larger model used. Gpt 2 can run on a pc.
--- End quote ---
Correct, the source code of GPT-2 and GPT-3 are exactly the same. The training data of GPT-3 is much larger and the resulting trained neural net is much larger. OpenAI published the source code, but not the training data or trained neural net.
There is a group at https://eleuther.ai who took the source code and created their own training dataset a little larger than GPT-3 with some improvements. They are currently training it and maybe by August will complete it and will publish both the training data and trained model.
--- End quote ---
In part the reason they need so much computational resource is that everything is in a high level language usually Python. I think resources could be more efficient if they used asembly. However thats only a part of it. The other part is that they are having the computer figure out everything or more than is nessicary. Think about how most people work for example... Its raining outside. You walk outside you get wet.... etc.,etc. after the figuring out phase for most people it just becomes a general rule of thumb... Its raining outside... is it raining and windy or just raining... if raining and windy get rain coat... if not windy get umbrella...other than varying levels of complexity you get the idea. What gpt needs to be able to do is bridge the gap from the training model to rule of thumb so ultimatley the gpt only really becomes active as the gazillion rules of thumb come up against something previously not covered by simpler rules creating new or refined rules to the existing database of rules.
I wanted to address the use of high level languages.. While Hal is probably devloped in Visual Studio, we must understand why? Lets look at why most operating systems and most software is written in higher level languages in the first place. Its not just about devlopment time but portability from computer to computer. Compiled languages like C are easier for humans to maintain, and an optimized C compiler is pretty tough to compete with. For example if you were to try to translate some C code to assembly you might well wind up with slower running assembly code than the compiled version. However if we start thinking how the machine thinks as so to speak we would look at things differently. Instead of looking at compilers and high level languages the way we do we might look at ms macro assembler or Hla (Randall Hyde) if you create macros within these assembly IDEs you can wind up with your high level language in a lower level form easier for the machine to optimize and can be made prortable. Macros being basically sections of reuseable code can also become part of the assemblers standard library. I know this idea is different but Im just looking at how we could restructure the way we do things to make it easier for the learning machine across different platforms. It would be nice to have a phone that could run a regular pc os quickly but as we have it now our not so smart phones boot up a version of linux then boots the android os in a virtual machine most of this written in a high level language that takes forever to bootup. However like many other systems the reason it is the way it is has more to do with commercial interests and politics than writing good code for a given purpose. Just look at the history of internet explorer as an example or how over the years more and more software is becoming subscription rentware to keep the money flowing and not really exploring the efficiency's and limits of more powerful modern computers.
Carl2:
When I first began thinking of GPT3 I thought about downloading the coding on to a storage device connected to the computer, SSD's do not contain enough memory so you would have to use a Hard drive in the tera byte range, cost is about $200 and I'm not sure if that would be large enough. It would eliminate paying for the cloud but it might make it possible to modify the training. At this time I can't even get a connected to a cloud with GPT 3. Looks like a lot of fun and You Tube has a lot of apps to do different things.
Carl2
lightspeed:
ROBERT , i don't know what is going on but i downloaded this from the site before as art told me to to get the hal changed over to the free no licensing version and it later gave me the 31 day trial is up error .
lightspeed:
art sent me the link again i went to it and started to download it , downloaded it and was going to install as administration and it showed this .as far as i know this is what is already installed . so why is it giving me the 31 days are up . i posted it elsewhere after it asked for my email and password and it said it is now the full version . i think when changing this over the hal window should display hal full free version . anyway i need to get this fixed . also when it said it was the free full version i noticed it had changed my off line use for hal back to cloud settings etc. and i had to change them back again. :P
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version