Android Question B4X with AI Machine Learning

[Global Robot Empire]

@Erel , can we Crowdfund you to incorporate AI into B4X – with the help of some Experts, if needed?

Not the Language Models – more like Deep Machine Learning to do interesting things.

Train the AI – like Tensor with Deepfakes.

But on a lower level, using APIs maybe – without the need for Nvidia Chips.

 

[BlueVision]

Personally, I see no reason to use the so-called AI in B4X. Why?

- Are the current functions of so-called nowadays ‘AI’ creative?
- Is today's AI capable of independently developing previously non-existent programme solutions that actually work?
- Does today's AI also consider the consequences of its work like a human does?

3x no.

All attempts on my part to incorporate currently available AI into programmes have so far failed. The result has always been unusable code.
Today's AI is perhaps good enough to quickly extract the right solution from existing solutions at breathtaking speed. But it is not creative. It is always just a copy of existing solutions, apart from a few highlights.
Seen in this light, it is actually still far too early to even talk about AI. Today's AI is a clever and super-fast search engine. Nothing more and nothing less.

Of course this could change, but not in the foreseeable future.

Then we would no longer need the wonderful B4X programming environment. Then the AI would provide us with the finished programme and nobody would be interested in how the whole thing works and was implemented. And that would be a huge problem, if not downright dangerous in certain areas.

At the moment, ‘AI’ is a good start with apparently good marketing.

 

[Magma]

Ofcourse this thread is not belong to B4A Questions... but may be at chit-chat... however... i think AI can help as a good Search-Machine for B4X and may be for small routines creating pseudocode.

I am sure that Erel already making experiments "years" now...

Do not forget that Erel is always steps ahead...

 

[Global Robot Empire]

Personally, I see no reason to use the so-called AI in B4X. Why?

- Are the current functions of so-called nowadays ‘AI’ creative?
- Is today's AI capable of independently developing previously non-existent programme solutions that actually work?
- Does today's AI also consider the consequences of its work like a human does?

3x no.

All attempts on my part to incorporate currently available AI into programmes have so far failed. The result has always been unusable code.
Today's AI is perhaps good enough to quickly extract the right solution from existing solutions at breathtaking speed. But it is not creative. It is always just a copy of existing solutions, apart from a few highlights.
Seen in this light, it is actually still far too early to even talk about AI. Today's AI is a clever and super-fast search engine. Nothing more and nothing less.

Of course this could change, but not in the foreseeable future.

Then we would no longer need the wonderful B4X programming environment. Then the AI would provide us with the finished programme and nobody would be interested in how the whole thing works and was implemented. And that would be a huge problem, if not downright dangerous in certain areas.

At the moment, ‘AI’ is a good start with apparently good marketing.

You are right. Agree completely

 

[ppgirl]

B4X can already do AI on the device side, such as Tensorflow lite. In addition, I also implemented pytorch_android_lite through inline java. B4X can achieve AI that Android can do.
Google's latest attempt to develop LLM and AIGC on Android. However, neither can produce a usable APK, because it is necessary to download a large model file separately, and the effect is far inferior to that of a PC with Nvidia GPU.

Find the right AI/ML solution for your app | Android Developers

developer.android.com

 

[MicroDrie]

i think AI can help as a good Search-Machine for B4X

I totally agree with this.

and may be for small routines creating pseudocode.

Indeed, the produced "pseudocode" is not always correct. AI can very quickly read a lot of factual documents, distance itself from those facts and sometimes come up with something surprising new reuse based on a different assumption. But as more creativity is needed to come up with something completely new and the amount of underlying facts becomes more limited, the applicability of the solution offered decreases.

 

[Tim Chapman]

I have used Perplexity AI to make B4X code. It gets it right about half the time.

 

[MicroDrie]

I have used Perplexity AI to make B4X code. It gets it right about half the time

You can tell from the answers given that AI has not (yet) fully understood the differences between B4X, B4A, B4J and B4I and newer versions, so that an answer contains an option that does not yet exist or no longer exists.

 

[Global Robot Empire]

AI Impact On IT Jobs

Imagine when Computers first appeared.

People said "Oh, this can't replace Humans."

AI is at the same stage.

But remember, what they give or show us is only about 1% of what TPTB have and are using.

 

[MicroDrie]

I think AI is a fantastic tool to draw a correct conclusion based on ever-increasing experience, such as cancer research from breast images. But I see similarities with a static approach, the larger the collection, the more accurate the conclusion. A creative approach is more based on letting go of everything, in order to look at it in a completely different way. But then AI lacks or has too small a collection of facts to give a statically sound answer. Then it seems that AI, based on the interpretation, seems to choose the statistically most likely choice. based on the question. Handy to be helped on your way if you have heard the bell tolling, but do not know exactly where the clapper is, as long as you realize that the answer may not necessarily be correct.

 

[Global Robot Empire]

As a learned Rabbi aptly commented:

AI SHMEYAI