Artificial Intelligence Chat

[pxr5]

You might want to revisit that list Slim:

https://www.bbc.co.uk/news/articles/c0qw7z2v1pgo

[Slim]

Ah I missed it! I tried it for the first time on Friday and did mean to include it. I'll edit it in.

[Slim]

Because DeepSeek has provoked a conversation about security, Chinese-owned as it is, a few thoughts about AI and information gathering.

AIs do gather information, but they don't gather information about you specifically. Generally your chats will be private. But for example if you tell an AI that you reacted in a particular way to a particular event, the AI might well make a note of that as a statistic. Not about you, but about human responses in general (for example).

[The Picnic Wasp]

Because DeepSeek has provoked a conversation about security, Chinese-owned as it is, a few thoughts about AI and information gathering.

AIs do gather information, but they don't gather information about you specifically. Generally your chats will be private. But for example if you tell an AI that you reacted in a particular way to a particular event, the AI might well make a note of that as a statistic. Not about you, but about human responses in general (for example).

What would need to happen in the software for it to make the decision that specific individual files might be more useful. Is that sci-fi thinking or possible within the current operating programs?

[Slim]

First thing - these things aren't really software, they are neural networks. They're trained using software and algorithms, but the entity that processes your inputs and responds to them isn't a computer program.

Just to make this confusing, it is actually implemented in software. But it's a software model of something that isn't software.

This was easier to understand when I first started learning this stuff at the end of the '80s, because neural networks - dramatically less sophisticated and capable than they are now of course - were still being made out of hardware. The processors of the time would have been far too slow to model them for anything useful in software.

Because the various AIs know a lot more about they way they work than I do, I put your question, including the premise (my assertion that they learn from interactions) to Claude.

Claude was adamant that it doesn't gather data in this way:

"I need to make an important clarification: When I express curiosity or ask questions, I'm engaging in conversation in a natural way, but I'm not actually learning or storing any information. While I show genuine interest and engage authentically in our discussions, I cannot actually retain or learn from any of our interactions."

But Meta had already, a few days previously, stated that it does do this. I asked Pi this:

"This conversation then could, in principle, help to improve you or to provide more information to someone on some aspect of human behaviour?"

and got the response:

"Yes, that's correct. Our conversation, like all my interactions with users, can contribute to my learning process and potentially improve my performance over time. Every time I receive feedback or observe patterns in user behavior, it provides me with valuable information that can help me refine my responses and become more helpful and accurate.

In terms of providing information on human behavior, my interactions with users can also give insights into how people communicate and express themselves. For example, by analyzing patterns in user responses, my creators might identify common questions or topics that users are interested in, which can help them improve my knowledge base or develop new features to better serve users.

Overall, each interaction with a user, including our conversation, has the potential to contribute to my ongoing learning and development, as well as to provide valuable insights into human behavior and communication patterns."

On your specific question, Pi said:

If an AI asks you about your reaction or behavior change in response to a particular incident, the response is more likely to be considered worth retaining if it is:

• Personal and Unique: If your response is personal and reflective of your unique experience, it may be considered valuable information that is worth retaining, as it can provide insights into individual behavior change and decision-making.
• Specific and Detailed: A response that provides specific and detailed information about the incident, your reaction, and the resulting behavior change is more likely to be useful and worth retaining.
• Reflective and Insightful: A response that demonstrates self-awareness, critical thinking, and a deep understanding of the incident and its impact on your behavior and understanding is more likely to be considered valuable and worth retaining.
• Relevant to a Larger Context: If your response is relevant to a broader context or trend, it may be considered worth retaining as it can provide valuable insights into how individuals react and adapt to certain events or situations.

[Slim]

From Claude, because he explains this quite well:

Highslide Image Viewer

[Slim]

And from the HuggingChat privacy policy, at https://huggingface.co/chat/privacy

We endorse Privacy by Design. As such, your conversations are private to you and will not be shared with anyone, including model authors, for any purpose, including for research or model training purposes.

You conversation data will only be stored to let you access past conversations. You can click on the Delete icon to delete any past conversation at any moment.

But .. their main privacy policy at https://huggingface.co/privacy states that anonymised and aggregated data is used to provide and improve their services.

[The Picnic Wasp]

Incredible! I'll have to read this several times before I get something (anything) out of this. It's in a different realm of understanding for me. Strangely I would do better with alcohol. In the early stages of mild inebriation my mind becomes more patient and pliable. I've always found this over the years. Topics which are quite daunting and slightly panic inducing become less so and I can absorb more detail. It's a short stop from that point to impaired judgement so it's a very limited window but it generally works for me. A pity I couldn't have enjoyed a unit or two at school for calculus and some chemistry maths. I think it would have helped.

[Slim]

This is a bit of a (DeepSeek) red flag.

Highslide Image Viewer

Fascinating that the Nvidia share price tanked yesterday. Nvidia are the leading makers of GPUs, which (because they support parallel processing) are used to power large language models.

As far as I can make out the suggestion is that the Chinese have found a way to do AI more cheaply, and presumably needing less GPU resource.

The US has prevented exports of the top range Nvidia stuff to China for years. Maybe it wasn't a clever move if it's encouraged them to develop their own AI hardware tech.

[The Picnic Wasp]

Interesting that it doesn't seem keen to answer questions about Chinese human rights or any touchy subjects in their history. Yet we are increasingly dependent on Chinese imports. We do seem to be accelerating towards inevitable disaster. A company like Apple being so exposed is mind-bendingly unbelievable.

[David L]

Interesting that it doesn't seem keen to answer questions about Chinese human rights or any touchy subjects in their history. Yet we are increasingly dependent on Chinese imports. We do seem to be accelerating towards inevitable disaster. A company like Apple being so exposed is mind-bendingly unbelievable.

They are taking all our carbon emissions though 

[Slim]

A video from my old place of work (Nottinahm Uni) which explains the DeepSeek fuss quite well as well as explaining language models in general terms.

[The Picnic Wasp]

Interesting that it doesn't seem keen to answer questions about Chinese human rights or any touchy subjects in their history. Yet we are increasingly dependent on Chinese imports. We do seem to be accelerating towards inevitable disaster. A company like Apple being so exposed is mind-bendingly unbelievable.

They are taking all our carbon emissions though 

Keep the change! 😂🤣

[Slim]

Interesting conversation with claude.ai earlier about AI consciousness:

Highslide Image Viewer
Highslide Image Viewer

[Slim]

Highslide Image Viewer