More than a thousand experts in the field of high technology and artificial intelligence have signed an open letter calling for the suspension of training of AI systems “more powerful than GPT-4”, the most popular and advanced neural network of this nature to date, for six months.
Among the signatories are researchers from DeepMind, Harvard, Oxford and Cambridge. As well as much more media personalities: Elon Musk, Apple co-founder Steve Wozniak, philanthropist Andrew Yang, Pinterest co-founder Evan Sharp and many others.
All of them believe that AI carries serious risks for society and humanity. In their opinion, powerful artificial intelligence systems should be developed only when they are sure that the effects will be positive.
During the six months for which the pause will be taken, the authors of the letter propose to develop common security protocols.
This must be done jointly with the authorities of states, it is necessary to form competent regulatory bodies and come up with systems of differences in products created by man and artificial intelligence.
Among the signatories are researchers from DeepMind, Harvard, Oxford and Cambridge. As well as much more media personalities: Elon Musk, Apple co-founder Steve Wozniak, philanthropist Andrew Yang, Pinterest co-founder Evan Sharp and many others.
All of them believe that AI carries serious risks for society and humanity. In their opinion, powerful artificial intelligence systems should be developed only when they are sure that the effects will be positive.
During the six months for which the pause will be taken, the authors of the letter propose to develop common security protocols.
This must be done jointly with the authorities of states, it is necessary to form competent regulatory bodies and come up with systems of differences in products created by man and artificial intelligence.
ChatGPT banned by regulator in Italy, which says there's no legal basis for using personal data to train the chatbot following data breach
https://www.businessinsider.com/cha...rs-in-italy-over-data-privacy-concerns-2023-3
I find this post very interesting https://www.diyaudio.com/community/threads/chatgpt-ideas-about-amplifier-design.394541/post-7243616
simply because if the gain stages are very transparent we end up listening to the output transistor and it seems each output transistor has a signature sound. Whats anyones take on the following output pairs. Describe perceived performance across the audio band
MJL4281/MJL4302
2SA1386A/2SC3519A
2SC5200/2SA1943
simply because if the gain stages are very transparent we end up listening to the output transistor and it seems each output transistor has a signature sound. Whats anyones take on the following output pairs. Describe perceived performance across the audio band
MJL4281/MJL4302
2SA1386A/2SC3519A
2SC5200/2SA1943
Douglas Self had done an article sometime back on TPC with interesting insights. However its also good to note there are many variations of frequency compensation and nothing is cast in stone on how to handle and shift dominant poles to a point where they cause no trouble
https://www.eetimes.com/distortion-...-vii-frequency-compensation-and-real-designs/
https://www.eetimes.com/distortion-...-vii-frequency-compensation-and-real-designs/
Heres Chat Gpts take
"
Yes, it's true that there are many variations of frequency compensation, and the specific approach taken will depend on the particular circuit design and its requirements. Frequency compensation is a critical aspect of amplifier design as it helps to ensure stability and prevent oscillation.
Douglas Self is a well-known authority in the field of audio amplifier design and his article on TPC (Transconductance-to-Current Conversion) provides some interesting insights into a particular compensation technique. However, as you mentioned, there are other compensation techniques that can also be used to address stability issues and shift dominant poles.
Some other common compensation techniques include Miller compensation, pole splitting, and nested Miller compensation. Each technique has its own advantages and disadvantages and may be more or less suitable depending on the specific application.
In any case, it's important to carefully analyze the stability of a circuit during the design process and choose the appropriate compensation technique to ensure that the amplifier operates correctly and reliably.
"
"
Yes, it's true that there are many variations of frequency compensation, and the specific approach taken will depend on the particular circuit design and its requirements. Frequency compensation is a critical aspect of amplifier design as it helps to ensure stability and prevent oscillation.
Douglas Self is a well-known authority in the field of audio amplifier design and his article on TPC (Transconductance-to-Current Conversion) provides some interesting insights into a particular compensation technique. However, as you mentioned, there are other compensation techniques that can also be used to address stability issues and shift dominant poles.
Some other common compensation techniques include Miller compensation, pole splitting, and nested Miller compensation. Each technique has its own advantages and disadvantages and may be more or less suitable depending on the specific application.
In any case, it's important to carefully analyze the stability of a circuit during the design process and choose the appropriate compensation technique to ensure that the amplifier operates correctly and reliably.
"
Is it you or the BOT that is talking? You I suppose. But the language is so well versed that I doubt it - but if wrong, congratulate you.
I suppose that if communicate a lot you will learn from the BOT and pick up its language skills an as such, probably improve yourself. This is nice 🙂
//
Switch, soft start, fuse - Protection.
In extreme cases, remove the plug from the outlet or pull the cord 🙁
So never. Its an absolute oxymoron. At least with the current trained with human input "AI". Perhaps future real AI could be let free as it will build a completely new knowledge base and might, if we are lucky, propose and induce a society scheme that is sustainable and make people in general happy. The current version is of course greedy - it has learnt from us. This is why Musk et al wants the future ones of the stage - navigating a greed saturated "killing field" is what they master.
//
Last edited:
Chat GPT still has a few kinks, for example in previous chat it extrapolated TPC (two pole compensation) wrong. Chat GPT gets these things wrong due to how it works. This approach of using educated guesses could generate new knowledge that may have been overlooked. Its akin to mining where the mineral is only a small percentage of the whole. In all the garbage it generates part of that garbage is gold.
"El-Mahdi El-Mhamdi, a former research scientist at Google, said he left the company in February over its refusal to engage with ethical AI issues head-on. Late last year, he said, he co-authored a paper that showed it was mathematically impossible for foundational AI models to be large, robust and remain privacy-preserving."
Bloomberg
Bloomberg
Yes.
It's pretty clear that any company providing such software is placing itself in clear legal jeopardy.
Not to mention the jeopardy of the user when acting on the provided misinformation.
It's pretty clear that any company providing such software is placing itself in clear legal jeopardy.
Not to mention the jeopardy of the user when acting on the provided misinformation.
If it "thinks" that is the best option.
No ethics, no morals, no regrets. What could go wrong?
On the other hand, people have morals and judgement.
https://www.vox.com/2018/9/26/17905796/nuclear-war-1983-stanislav-petrov-soviet-union
No ethics, no morals, no regrets. What could go wrong?
On the other hand, people have morals and judgement.
https://www.vox.com/2018/9/26/17905796/nuclear-war-1983-stanislav-petrov-soviet-union
Last edited:
Also of possible interest for some folks:
https://proceedings.neurips.cc/paper/2017/file/f4b9ec30ad9f68f89b29639786cb62ef-Paper.pdf
https://arxiv.org/pdf/1802.07927.pdf
Basic gradient descent is used for all sorts of purposes nowadays. There are problems that have no deterministic solution, instead the only possible approach may be to search a solution space for a local optimum.
https://proceedings.neurips.cc/paper/2017/file/f4b9ec30ad9f68f89b29639786cb62ef-Paper.pdf
https://arxiv.org/pdf/1802.07927.pdf
Basic gradient descent is used for all sorts of purposes nowadays. There are problems that have no deterministic solution, instead the only possible approach may be to search a solution space for a local optimum.
Last edited: