A couple of days ago I started studying OpenAI's APIs with the goal of creating a Hyperlambda machine learning algorithm that produces perfect Hyperlambda code given natural text as its only input. Today I released the first BETA release of my research and development, and I'm already at an accuracy level of roughly 90%. Watch the following video for a demonstration.
How it works
OpenAI allows you to start out with one pre-defined model or "engine". Then you can provide training material to the engine, which are questions with answers. The more data you provide, the higher the accuracy of the result later becomes. With less than 700 snippets of Hyperlambda code associated with prompt questions, I was able to teach OpenAI most of Hyperlambda's syntax, bringing accuracy up to roughly 90%. Notice, this was 2 days of work for me. I obviously intend to continue improving upon it, teaching it more in the near future. But already it can produce Hyperlambda code for questions such as ...
Create a Hyperlambda endpoint that inserts name, address and zip into my customers table in my crm database
And the code it produces given the above question is as follows ...
.arguments
name:string
address:string
city:string
country:string
data.connect:magic
data.create
table:customers
values
name:x:@.arguments/*/name
address:x:@.arguments/*/address
city:x:@.arguments/*/city
country:x:@.arguments/*/country
response.status.set:201
return
result:Insert was a success
The above code is roughly 90% accurate, but I believe I can bring it beyond 99% accuracy in a couple of weeks, effectively rendering Hyperlambda "obsolete", or rather the need to understand Hyperlambda "obsolete", since you can use a humanly friendly text prompt in Hyper IDE to have artificial intelligence automatically produce working backend code for you. Another example of code it produced for me can be found below ...
Send an email to John Doe having en email address of john@doe.com
The result of the above was ...
mail.smtp.send
message
to
.
name:John Doe
email:john@doe.com
from
.
name:Jane Doe
email:jane@doe.com
subject:This is the email subject
entity:text/plain
content:This is the email body
The above is even more accurate, in fact it's 100% perfectly accurate, ignoring the fact that I didn't provide a "from" part in my question to OpenAI. But that was because I didn't phrase my question correctly.
If you want to try it out, you can register for an AI based Low-Code Aista Magic Cloudlet below.
Top comments (6)
omg stop there won't be any jobs left 😂
Hehe :D
I am over worked, I need an assistant ;)
What is your opinion on Rust? i'm thinking of switching to it from Typescript.
Hehe, from TypeScript to Rust ...? :D
I have no idea to be honest with you. I would presume Rust and TypeScript are two different domains in regards to what problems to solve, as in Rust being low level alternative to C++ while TypeScript is mostly a high level alternative to JavaScript. But I've got no idea. I suspect it's much better than C++, which arguably doesn't take much though ... :/
Is Rust another mass psychosis or is it an ok lang? I saw bunch of people switching to it and i wanted to try it too.
Everything that is based upon classes tying the subject to the verb is to some degree the wrong direction, for reasons I outline in my original article. However, at this point, I suspect you just have to pick your madness. I suspect on the scale of madness, C++ is "batshit straight jacket crazy", where Rust is "being a little but neurotic on Mondays" ...
It took us 2,000 years as a specie to for the most parts agree upon that people couldn't walk on water. I suspect it'll take us the same amount of time realising OO was a dead end ...
You might as well go with the flow. Less painful that way ... :/