Chat GPT: WOW.

[raif71]

Elon Musk and others urge AI pause, citing 'risks to society'

Elon Musk and a group of artificial intelligence experts and industry executives are calling for a six-month pause in developing systems more powerful than OpenAI's newly launched GPT-4, in an open letter citing potential risks to society.

www.reuters.com

 

[Martin]

Unfortunately, for unscrupulous college students you can feed a document into an AI detector program and it will tell you whether it was generated by AI or not.

One of the best uses for AI is writing computer routines. Give AI the inputs, outputs, and parameters for a routine and it can spit out bug free code in seconds. A little additional tweaking and it will correctly detect unexpected input, handle boundary cases, and throw/handle exceptions.

One of the hottest fields in computer science is AI prompt generation.

Martin

 

[NoxMorbis]

No, it much worse. It will just make stuff up also when totally not appropriate.

I asked it to list a few software libraries to do a specific function, and the first one it came up with was totally fictional.

Once confronted it made some empty apologies

lol -- interesting. As always, confirm.

 

[NoxMorbis]

Chat GPT is better for brainstorming content ideas than asking data-driven questions.

For sure, and I already knew the answers to my questions, so I could tell when it was off the mark. Mostly it was because my questions were not concise enough.

 

[NoxMorbis]

Elon Musk and others urge AI pause, citing 'risks to society'

Elon Musk and a group of artificial intelligence experts and industry executives are calling for a six-month pause in developing systems more powerful than OpenAI's newly launched GPT-4, in an open letter citing potential risks to society.

www.reuters.com

Of course he does because he has competing interest that isn't where Chat GPT is. He wants time to catch up. I trust Elon Musk about as much as zero. He, himself, is the personification of AI gone bad. The guy is just another predator.

 

[raif71]

Of course he does becxsaue he has competing interest that isn't where Chat GPT is. he wants time to catch up. I trust Elon Musk about as much as zero. He, himself, is the personification of AI gone bad. The guy is just another predator.

Owh I thought that he just got word from the future that robots ruled the earth and he wants to stop it now from happening

 

[solderdude]

ME: Does the Texas instruments 3255 amplifier chip have built in spike protection for clipping?

Data sheet states there is an output pin called CLIP_OTW (pin 21) which goes low on detection of near clipping condition (typically > 0.01% dist).
That can be used to manipulate the input signal externally.
The amp will just clip when overloaded (above 2V input) but does it in a controlled way by not sticking to a voltage rail but maintaining switching action.

and DC speaker protection

This, however, is not DC voltage protection but only a DC current protection for when the DC current between the BTL outputs exceed 1.5A.
The input is AC coupled (10uF) and should be applied by the designer.

Re: ChatGPT: It can give anything between accurate info and nonsense.
That said, when asking "audio gear" designers you can also get anything between accurate info and nonsense.
This includes my technical interpretation of the TPA3255 datasheet as I never played with that chip

Verification is always key.

 

[Peterinvan]

So I asked ChatGPT for different opinions on the Meze 109 Pro.
Here are its first and second (refreshed) answers. I looks to me like if we loose audio writers to ChatGPT, we will not have any source material to report:


The Meze 109 Pro headphones have generally received positive reviews from audio enthusiasts and critics. However, as with any product, there are differing opinions and experiences among users.
Here are some common positive and negative opinions on the Meze 109 Pro headphones:
Positive opinions:
The sound quality is excellent, with a balanced and detailed sound signature that works well across different genres of music.
The headphones are well-built with high-quality materials and feel comfortable to wear for extended periods of time.
The included accessories, such as the hard case and multiple cables, add value to the overall package.
The headphones have a sleek and stylish design that many users appreciate.

Negative opinions:
Some users find the bass to be lacking, particularly in comparison to other headphones in a similar price range.
The headphones may not fit everyone's head comfortably, as the ear cups have a relatively small opening and may not accommodate larger ears.
The headphones are not noise-cancelling, so they may not be the best choice for use in noisy environments.
The price point may be considered high for some users, especially if they are not audiophiles or don't prioritize high-end audio equipment.

Overall, the Meze 109 Pro headphones are generally well-regarded for their sound quality and build quality. However, whether they are the right choice for an individual will depend on their personal preferences and priorities.

REFRESH:

"As an AI language model, I don't have personal opinions, but I can provide you with some information on the Meze 109 Pro headphones and different opinions from various sources.
The Meze 109 Pro headphones are over-ear headphones designed for audiophiles who want a high-quality listening experience. They feature a sleek, minimalist design with a sturdy metal frame and soft, comfortable earpads.
Some reviewers have praised the Meze 109 Pro headphones for their impressive sound quality, with a detailed, balanced sound that doesn't overemphasize any particular frequencies. They also note that the headphones have a wide soundstage, which creates a sense of depth and immersion in the music.
Other reviewers have criticized the headphones for their fit, noting that they can be uncomfortable for some users, especially those with larger heads. Some users have also noted that the headphones can feel heavy and clamp down too tightly on the head.
In terms of pricing, the Meze 109 Pro headphones are considered a mid-range option, with a price point around $400. Some reviewers feel that the price is reasonable for the quality of the headphones, while others think that there are better options available at a similar price point.
Overall, the Meze 109 Pro headphones have received mostly positive reviews, with many users praising their sound quality and design. However, as with any headphones, there are varying opinions on their fit and comfort, so it's important to try them out for yourself before making a purchase decision.

 

[NoxMorbis]

Data sheet states there is an output pin called CLIP_OTW (pin 21) which goes low on detection of near clipping condition (typically > 0.01% dist).
That can be used to manipulate the input signal externally.
The amp will just clip when overloaded (above 2V input) but does it in a controlled way by not sticking to a voltage rail but maintaining switching action.



This, however, is not DC voltage protection but only a DC current protection for when the DC current between the BTL outputs exceed 1.5A.
The input is AC coupled (10uF) and should be applied by the designer.

Re: ChatGPT: It can give anything between accurate info and nonsense.
That said, when asking "audio gear" designers you can also get anything between accurate info and nonsense.
This includes my technical interpretation of the TPA3255 datasheet as I never played with that chip

Verification is always key.

So was it in the ballpark?

 

[voodooless]

So was it in the ballpark?

No!

 

[Jim Shaw]

I see the real power of AI as being able [eventually] to consider more information resources than most human researchers would. The real weakness could be skewed or prejudicial results that AI might believe are truths. In the foreseeable future, AI probably doesn't have much ability to ascertain the truth -- just like humans don't.

 

[kemmler3D]

It's simply an interface for a repository of information, and if you give it unclear instructions, it will give you erroneous output.

It will actually give you erroneous output under many conditions.

ChatGPT and other LLMS are excellent at producing output that resembles the thing you want. In fact, that's all they can do. Because they are good at this, the output is often what you actually do want.

However, it's currently pretty bad for fact-based output, because in order to know whether it's delivered the facts correctly, you already need to know what those facts are. It often makes things up, and even worse, mixes made-up and real things in the same output. Most of the time it would be faster and easier to write it yourself.

 

[solderdude]

CHATGPT: Yes, the Texas Instruments TPA3255 amplifier chip does have built-in spike protection to prevent clipping and potential damage to the amplifier or speakers.
The TPA3255 chip includes a range of protection features, including over-current protection, over-temperature protection, and DC speaker protection. In addition, the chip has a built-in limiter to prevent clipping by dynamically adjusting the gain of the amplifier in response to the input signal level. The limiter works by reducing the gain of the amplifier when the input signal level exceeds a certain threshold, preventing the signal from reaching the maximum output level of the amplifier. The TPA3255 chip also includes a range of diagnostic and monitoring features, including DC offset detection, fault reporting, and thermal monitoring, which can help to improve the reliability and safety of the amplifier. Overall, the TPA3255 chip is designed to provide high-quality audio performance while ensuring the safety and protection of the amplifier and speakers.

So was it in the ballpark?

The correct info is made green, the incorrect parts red and the incomplete parts that could be interpreted incorrectly in blue.

 

[xaviescacs]

Chat GPT is a bigmouth.

 

[NoxMorbis]

No!

So like I accused, it's really not any AI, just a librarian language Ai that can talk like a human, but has no ability to reason at all. Very disappointing. It's just an interface that sounds human but is worse than humans at searching the web and making logical inferences about the information. lol Sounds about right for Microsoft.

This ain't no Star Trek computer.

 

[NoxMorbis]

It will actually give you erroneous output under many conditions.

ChatGPT and other LLMS are excellent at producing output that resembles the thing you want. In fact, that's all they can do. Because they are good at this, the output is often what you actually do want.

However, it's currently pretty bad for fact-based output, because in order to know whether it's delivered the facts correctly, you already need to know what those facts are. It often makes things up, and even worse, mixes made-up and real things in the same output. Most of the time it would be faster and easier to write it yourself.

This is true. Since I do know how to go about finding factual information, I can tell when it is inaccurate or flubbing, and if I don't, I know enough to double check, like the bit it spit out about the TI 3255 amp and spike protections. Like you said, it's good at spitting out a well seritten "X" as long as you know how to fact check it beofor using it: e.g., a table of contents or summary of a manual.

 

[NoxMorbis]

The correct info is made green, the incorrect parts red and the incomplete parts that could be interpreted incorrectly in blue.

Hey that was really nice of you to do. Very good. So the Texas Instruments TPA3255 amplifier chip REALLY DOES NOT have built-in spike protection to prevent clipping?

 

[NoxMorbis]

ChatGPT is, in my opinion, and excuse my language FUCKING GARBAGE for finding factual content.

 

[Elitzur–Vaidman]

ChatGPT is, in my opinion, and excuse my language FUCKING GARBAGE for finding factual content.

Are you using GPT3.5 or are you paying for GPT 4?

 

[Elitzur–Vaidman]

Using GPT-4:
Me: Does the Texas instruments 3255 amplifier chip have built in spike protection for clipping?
ChatGPT:

As an AI language model, I do not have real-time information about specific electronic components that may have been released after my knowledge cutoff date in September 2021. Texas Instruments 3255 amplifier chip is not a known component within my knowledge base.

If you are referring to an operational amplifier or audio amplifier from Texas Instruments, it would be best to consult the datasheet for the specific model you are referring to in order to find information about built-in spike protection or other protection mechanisms against clipping.

If you are unsure about the specific chip model, I suggest visiting the Texas Instruments website and searching for the amplifier chip by its key specifications, or even contacting their support team for more information.