AI endeavours

[TUK020]

The idea that Tesla are in any way ahead is risible.

The idea that making Tesla's precocious cruise control autonomous under all conditions is going to be easier than expanding a geo-fence, or reducing the requirements for that geo-fence is risibler.

Optimus is just another of Musk's attempts to distract from the fact that the Roadster, The CyberTruck, The Semi and a version of FSD that is actually capable of autonomous driving are nowhere in sight, despite all being many years overdue.

Then that makes Musk a pretty smart guy, because his distractions are fooling a lot of people, including Ody the Risiblessed. And calls about the dangers of A.I. are more of the same.

Musk is a brilliant man, and like most brilliant people, is easily bored. He spotted a gap in the market and has exploited it brilliantly with Tesla. Not sure I would want to be a holder of Tesla stock now, when it is at last in the cross hairs of Chinese tackling his vertical integration cost advantage, or the 'legacy' manufacturers who are capable of attending to quality and customer satisfaction in volume production.
Meantime, Musk has gone on to achieve new brilliant things with the building competitive edge in SpaceX and Starlink. I suspect this cycle will continue.

[BobbyD]

The idea that Tesla are in any way ahead is risible.

The idea that making Tesla's precocious cruise control autonomous under all conditions is going to be easier than expanding a geo-fence, or reducing the requirements for that geo-fence is risibler.

Optimus is just another of Musk's attempts to distract from the fact that the Roadster, The CyberTruck, The Semi and a version of FSD that is actually capable of autonomous driving are nowhere in sight, despite all being many years overdue.

Then that makes Musk a pretty smart guy, because his distractions are fooling a lot of people, including Ody the Risiblessed. And calls about the dangers of A.I. are more of the same.

Musk is a brilliant man, and like most brilliant people, is easily bored. He spotted a gap in the market and has exploited it brilliantly with Tesla. Not sure I would want to be a holder of Tesla stock now, when it is at last in the cross hairs of Chinese tackling his vertical integration cost advantage, or the 'legacy' manufacturers who are capable of attending to quality and customer satisfaction in volume production.
Meantime, Musk has gone on to achieve new brilliant things with the building competitive edge in SpaceX and Starlink. I suspect this cycle will continue.

Credit where credit is due Musk is the P.T. Barnum of our age.

It's no coincidence that Space X is the company he isn't actually allowed to run because firing large bombs in to the sky is considered a bit dangerous.

If Gwynne Shotwell were running Tesla rather tham Space X I'd actually consider buying stock.

[odysseus2000]

Hollywood begins to realise what AI can do, just like weavers realized what Arkwright’s water frame could & did do (2 mins 59):

https://youtu.be/BW7Yu1wszo4

Regards,

[odysseus2000]

AI by 2030 could contribute more to the global than the combined output of China & India:

https://twitter.com/smartertrader/statu ... DCpgdbFBxg

Regards,

[odysseus2000]

31 minutes of what AI may mean to the world:

https://youtu.be/ciX_iFGyS0M

Regards,

[Sorcery]

31 minutes of what AI may mean to the world:

https://youtu.be/ciX_iFGyS0M

Regards,

The video is hopelessly optimistic about AI. ChatGPT4 get's it wrong very often. How on earth can anyone trust it with a medical diagnosis, speccing out the materials and sizes required to build a bridge to carry a specified load, or even program.
In its current version I would not trust it at all. viewtopic.php?f=83&t=37024#p552453

The interviewee said the job of software engineering would be gone in 5 years, which might be true but it won't be ChatGPT4. ChatGPT4 said to me it could not test code, which is a huge disadvantage when it's as buggy as it is.

My opinion hasn't changed since this post: viewtopic.php?p=592311#p592311

[ReformedCharacter]

31 minutes of what AI may mean to the world:

https://youtu.be/ciX_iFGyS0M

Regards,

The video is hopelessly optimistic about AI. ChatGPT4 get's it wrong very often. How on earth can anyone trust it with a medical diagnosis, speccing out the materials and sizes required to build a bridge to carry a specified load, or even program.
In its current version I would not trust it at all. viewtopic.php?f=83&t=37024#p552453

The interviewee said the job of software engineering would be gone in 5 years, which might be true but it won't be ChatGPT4. ChatGPT4 said to me it could not test code, which is a huge disadvantage when it's as buggy as it is.

My opinion hasn't changed since this post: viewtopic.php?p=592311#p592311

I agree about the shortcoming of ChatGPT4 but ChatGPT4 is only one type of AI, and I expect the improvements in AI to follow something like Moore's law in the rate of progress. This is just the beginning and there is a lot of money and resources being spent on it.

RC

[ReformedCharacter]

It's no coincidence that Space X is the company he isn't actually allowed to run because firing large bombs in to the sky is considered a bit dangerous.

If Gwynne Shotwell were running Tesla rather tham Space X I'd actually consider buying stock.

If sure GW is a highly capable manager and CEO but I get the impression - seeing interviews with Musk at Boca Chica - that he is very much involved with the engineering side of the business and the ethos of rapid iteration and innovative construction techniques. I think he calls the shots in that respect.

RC

[odysseus2000]

Interesting commentary by Ark on us drone delivery (19:00 to 27 minutes):

https://youtu.be/LoZmmCX5GuQ?si=xUPmsrX91Y5OK1JK

These are still baby steps, but the whole concept looks like it will be deployed. Zip wire who have big operations in Rwanda still seem to be a leader in this technology.

Regards,

[odysseus2000]

Super interesting discussion by an Ark team on why Bitcoin is the currency for AI (Just over 1 hour):

https://youtu.be/IQrRO6FoYl0?si=IOX8N6Ij33-fOSc9

This is a complex discussion & anyone not upto speed on AI & bit coin will need to do some research to get the most out of this.

The basic tenant is that AI can seek information for review by a human, improving knowledge work efficiency, with the AI agents paying for what they need using Bitcoin & then converting that to $. Savings look to be enormous, but the issue of garbage in/ garbage out still applies.

Regards,

[ReformedCharacter]

Meta, in partnership with Microsoft, seem to be making progress with AI and have made their product open source (Llama 2):

https://about.fb.com/news/2023/07/llama-2/

Also, Code Llama:

Code Llama is a code-specialized version of Llama 2 that was created by further training Llama 2 on its code-specific datasets, sampling more data from that same dataset for longer. Essentially, Code Llama features enhanced coding capabilities. It can generate code and natural language about code, from both code and natural language prompts (e.g., “Write me a function that outputs the fibonacci sequence”). It can also be used for code completion and debugging. It supports many of the most popular programming languages used today, including Python, C++, Java, PHP, Typescript (Javascript), C#, Bash and more.

RC

[odysseus2000]

Meta, in partnership with Microsoft, seem to be making progress with AI and have made their product open source (Llama 2):

https://about.fb.com/news/2023/07/llama-2/

Also, Code Llama:

Code Llama is a code-specialized version of Llama 2 that was created by further training Llama 2 on its code-specific datasets, sampling more data from that same dataset for longer. Essentially, Code Llama features enhanced coding capabilities. It can generate code and natural language about code, from both code and natural language prompts (e.g., “Write me a function that outputs the fibonacci sequence”). It can also be used for code completion and debugging. It supports many of the most popular programming languages used today, including Python, C++, Java, PHP, Typescript (Javascript), C#, Bash and more.

RC

Thank you!

It is interesting that Msft who paid Open-AI a fortune are now giving away Llama via Meta, while at the same time Copilot is $30 per month when ever it launches. As I understand it, Copilot was announced in March of 2023 but it still not available, giving some idea of the difficulty of launching these AI models for commercial applications.

I might try Llama when next I have to write some Arduino code.

Regards,

[ReformedCharacter]

It is interesting that Msft who paid Open-AI a fortune are now giving away Llama via Meta, while at the same time Copilot is $30 per month when ever it launches. As I understand it, Copilot was announced in March of 2023 but it still not available, giving some idea of the difficulty of launching these AI models for commercial applications.

I might try Llama when next I have to write some Arduino code.

Regards,

This might be a good place to start, I gave it a specification for a bash script that I had recently written and the generated code looked fine, only a simple test admittedly:

https://huggingface.co/spaces/codellama/codellama-13b-chat

RC

[odysseus2000]

It is interesting that Msft who paid Open-AI a fortune are now giving away Llama via Meta, while at the same time Copilot is $30 per month when ever it launches. As I understand it, Copilot was announced in March of 2023 but it still not available, giving some idea of the difficulty of launching these AI models for commercial applications.

I might try Llama when next I have to write some Arduino code.

Regards,

This might be a good place to start, I gave it a specification for a bash script that I had recently written and the generated code looked fine, only a simple test admittedly:

https://huggingface.co/spaces/codellama/codellama-13b-chat

RC

Interesting, Thank you!

Regards,

[Itsallaguess]

‎
Boston Dynamics apply Chat AI models to some of their robots -

https://www.youtube.com/watch?v=djzOBZUFzTw

Cheers,

Itsallaguess

[odysseus2000]

‎
Boston Dynamics apply Chat AI models to some of their robots -

https://www.youtube.com/watch?v=djzOBZUFzTw

Cheers,

Itsallaguess

So Marvin the paranoid android is born!

Great video giving an insight into what Optimus might do with its AI built in from the beginning, not added as an after thought.

Regards,

[odysseus2000]

Have we reached AGI or is this baloney?:

https://www.reuters.com/technology/sam- ... 023-11-22/

Does anyone with advanced knowledge of this field know if this is for real or just hype?

Regards,

[odysseus2000]

This 15 minute video does a good summary of the runners & riders & is very optimistic as to what this technology can do, perhaps too enthusiastic as of yet there have not been many huge advances, save for in Chess & Go, but interesting:

https://youtu.be/Hz8yZUXsYrs?si=FSE62mpJhezzU3zb

Regards,

[ReformedCharacter]

How Neural Networks Learned to Talk | ChatGPT: A 30 Year History

This video explores the journey of language models, from their modest beginnings through the development of OpenAI's GPT models & hints at Q*. Our journey takes us through the key moments in neural network research involved in next word prediction. We delve into the early experiments with tiny language models in the 1980s, highlighting significant contributions by researchers like Jordan, who introduced Recurrent Neural Networks, and Elman, whose work on learning word boundaries revolutionized our understanding of language processing. Featuring Noam Chomsky Douglas Hofstadter Michael I. Jordan Jeffrey Elman Geoffrey Hinton Ilya Sutskever Andrej Karpathy Yann LeCun and more. (Sam altman)

https://www.youtube.com/watch?v=OFS90-FX6pg

RC

[odysseus2000]

How Neural Networks Learned to Talk | ChatGPT: A 30 Year History

This video explores the journey of language models, from their modest beginnings through the development of OpenAI's GPT models & hints at Q*. Our journey takes us through the key moments in neural network research involved in next word prediction. We delve into the early experiments with tiny language models in the 1980s, highlighting significant contributions by researchers like Jordan, who introduced Recurrent Neural Networks, and Elman, whose work on learning word boundaries revolutionized our understanding of language processing. Featuring Noam Chomsky Douglas Hofstadter Michael I. Jordan Jeffrey Elman Geoffrey Hinton Ilya Sutskever Andrej Karpathy Yann LeCun and more. (Sam altman)

https://www.youtube.com/watch?v=OFS90-FX6pg

RC

Super interesting in that they did give far more information on how these models are configured and trained, or how they think they are.

What I find hard is that the relatively simple model described in the video does not seem to have the power of Alpha-Go which, as world Go and Chess champion, as some ability to either think or do something equivalent and to do so over relatively long periods without loss of piece position and to do so simultaneously against many opponents. There seems to be an entire structure of compute and deductive reasoning that is not being discussed.


Or am I missing something?

Regards,