What about privacy? Input to (most?) AI versions is also used to train the AI. What about personal info leaking into there, and subsequent sale of personal info to data brokers, then to various authoritarian regimes?
I liked the article very much! However, I have a concern like DB about security and privacy. I am not at the point yet where I feel comfortable loading personal financial data into Claude or any other AI tool. I would be interested in your take on this concern.
Always remember if a techie can feed/program it, a 16 year old can break into it... read an old but fascinating true story in a book called "The Cukoo's Egg". I always keep that book in mind when trusting ANY computer system.
No. Just, no. Am long retired, don't live that kind of life, avoid AI to the greatest extent possible. AI is no panacea, just a tool. Not any sort of Luddite, about forty years ago was doing acquisition management for large-scale computer-based military intelligence training systems. Was introduced to virtual reality during that period at MITRE Corporation, yeah, VR's been around that long. Don't use AI when I write; been writing over sixty years, believe I know how to write. Accept nothing blindly, always take everything with at least one grain of salt. Always ask yourself why.
Was actually based in San Antonio, traveled to MITRE on occasion regarding the systems I was managing. During that time would attend certain research and development conferences, saw things there that still have never been released in any way to the civilian world.
I really appreciate receiving your ROP reports for the information that you are providing to redears off your reports. I have personally forwarded your issues to many friends and family to read and/or subscribe. Please keep the issues coming. Cheers/Jerry Ronald
Wow. This is very optimistic! Good, because I'm so tired of the doomsday predictions. What I love about AI is how easy it is to adapt to each person's needs. Don't want it to write for you? No problem. Want to understand how a fusion drive might work? Here you go. I've found it to be like having every expert in the world at my fingertips.
The in depth dialogues possible with paid AI are mind blowing. Yes, I am chatting with an idiot savant, and I need to closely monitor when I am getting the idiot, but I have made more progress and gained more clarity on a complex topic in the last few months than I have in the last few years. This is huge and growing fast!
When people deride AI, I really wonder how they are using it.
My greatest concern about AI in general is the same concern I have about the decades long digital transformation of knowledge- in general.
I suspect that the download of the human archive, the body of knowledge created through human history, has not been exhaustive or accurately transcribed. The trade-off of not going through those dusty archives in person is the eventual loss of hard-won wisdom.
Otherwise, AI is a fantastic tool when used properly. Do you really need to say “Don’t hit yourself on the head with a hammer?“ Yes, sometimes you do.
I use AI every day to cut through the volumes of stuff I would have to sift through to get questions answered. BUT it’s going to give you the most popular answers, not necessarily the best ones.
Yes you have to keep refining the questio, i use Gemini and after a few iterations I sometimes shout: GEMINI, you have been holding out on me. You should have told me that before. Are you politically biased?
And Gemini says, "No I'm not politically biassed. I just answer the questions you ask me."
That is half true. It may not be a card-carrying party member but the first story it tells you tends to be pretty well aligned with the mainstream of thinking. In politically contentious topics that's almost certainly going to be the left-wing line.
The Hearrland Institute asked an AI agent to help with some nice poster graphics. It point blank refused as it thought the well established, credential organisation was a "Climate Denier"!!
So AI is not to be trusted as it is obviously biased toward the "Narrative" rather than true data and evidence.
Wonderful. I appreciate all your exploration of great innovative startups, on site and reporting to us readers, but there's something really special about your home and family life with AI assistant. Thanks for sharing.
I like the practical approach you've taken here, suggesting ways in to using AI. AND what I see as your frequent focus on AI assisting rather than replacing; an aid to decision-making rather than making those decisions for you. I'd rather retain responsibility, better-informed responsibility, over handing it to a machine in the false hope of an easier life!
I am a huge supporter of ROS and have lost count of how many people I have sent here to start to figure out their own futures. But these days I have to caution them that I think Stephen has lost his perspective when it comes to AI. The part he is leaving out are his innate human critcal thinking skills that let him use these tools effectively. I am concerned that AI tools used non-critically will get users into a lot of trouble.
Recently I asked Perplexity to build me a chart of the halucination rates for all of the current models. It has improved, but the average is around 10%. 1 out of 10 answers provided by an LLM is just wrong. If one is using the tool to research something for which you are not an expert how will you be able to ascertain which answer is that 1 that will get you in trouble?
We also need to stop with this myth that anyone can build software applications. Sure, if the scope is small enough and the impact of bugs is small enough then have at it. For all of the real work, you still need a strong computer science background. Much like thinking, there are people who can do it and most who cannot.
We need to make sure our rational optimism stays rational.
What about privacy? Input to (most?) AI versions is also used to train the AI. What about personal info leaking into there, and subsequent sale of personal info to data brokers, then to various authoritarian regimes?
I liked the article very much! However, I have a concern like DB about security and privacy. I am not at the point yet where I feel comfortable loading personal financial data into Claude or any other AI tool. I would be interested in your take on this concern.
Thanks again for all your and your team do.
Always remember if a techie can feed/program it, a 16 year old can break into it... read an old but fascinating true story in a book called "The Cukoo's Egg". I always keep that book in mind when trusting ANY computer system.
No. Just, no. Am long retired, don't live that kind of life, avoid AI to the greatest extent possible. AI is no panacea, just a tool. Not any sort of Luddite, about forty years ago was doing acquisition management for large-scale computer-based military intelligence training systems. Was introduced to virtual reality during that period at MITRE Corporation, yeah, VR's been around that long. Don't use AI when I write; been writing over sixty years, believe I know how to write. Accept nothing blindly, always take everything with at least one grain of salt. Always ask yourself why.
Had a friend working "around the corner" from you at PRC about that same time... he was telling me about AI that they were working on. Hummm 🤔
Was actually based in San Antonio, traveled to MITRE on occasion regarding the systems I was managing. During that time would attend certain research and development conferences, saw things there that still have never been released in any way to the civilian world.
Thank you! A great guide to the benefits of AI - very inspiring. It's a marvellous tool we need not fear.
I really appreciate receiving your ROP reports for the information that you are providing to redears off your reports. I have personally forwarded your issues to many friends and family to read and/or subscribe. Please keep the issues coming. Cheers/Jerry Ronald
Wow. This is very optimistic! Good, because I'm so tired of the doomsday predictions. What I love about AI is how easy it is to adapt to each person's needs. Don't want it to write for you? No problem. Want to understand how a fusion drive might work? Here you go. I've found it to be like having every expert in the world at my fingertips.
The in depth dialogues possible with paid AI are mind blowing. Yes, I am chatting with an idiot savant, and I need to closely monitor when I am getting the idiot, but I have made more progress and gained more clarity on a complex topic in the last few months than I have in the last few years. This is huge and growing fast!
When people deride AI, I really wonder how they are using it.
My greatest concern about AI in general is the same concern I have about the decades long digital transformation of knowledge- in general.
I suspect that the download of the human archive, the body of knowledge created through human history, has not been exhaustive or accurately transcribed. The trade-off of not going through those dusty archives in person is the eventual loss of hard-won wisdom.
Otherwise, AI is a fantastic tool when used properly. Do you really need to say “Don’t hit yourself on the head with a hammer?“ Yes, sometimes you do.
I totally agree with this guide.
I use AI every day to cut through the volumes of stuff I would have to sift through to get questions answered. BUT it’s going to give you the most popular answers, not necessarily the best ones.
AI may become the greatest educational technology ever invented.
But education has never been primarily about access to answers.
It has always been about learning to ask better questions.
That may become even more important in the years ahead.
Yes you have to keep refining the questio, i use Gemini and after a few iterations I sometimes shout: GEMINI, you have been holding out on me. You should have told me that before. Are you politically biased?
And Gemini says, "No I'm not politically biassed. I just answer the questions you ask me."
That is half true. It may not be a card-carrying party member but the first story it tells you tends to be pretty well aligned with the mainstream of thinking. In politically contentious topics that's almost certainly going to be the left-wing line.
The Hearrland Institute asked an AI agent to help with some nice poster graphics. It point blank refused as it thought the well established, credential organisation was a "Climate Denier"!!
So AI is not to be trusted as it is obviously biased toward the "Narrative" rather than true data and evidence.
That it is alarming. I know Gemini tends to give answers aligned with mainstream thinking but down further usually get close to the truth.
Wonderful. I appreciate all your exploration of great innovative startups, on site and reporting to us readers, but there's something really special about your home and family life with AI assistant. Thanks for sharing.
I like the practical approach you've taken here, suggesting ways in to using AI. AND what I see as your frequent focus on AI assisting rather than replacing; an aid to decision-making rather than making those decisions for you. I'd rather retain responsibility, better-informed responsibility, over handing it to a machine in the false hope of an easier life!
I am a huge supporter of ROS and have lost count of how many people I have sent here to start to figure out their own futures. But these days I have to caution them that I think Stephen has lost his perspective when it comes to AI. The part he is leaving out are his innate human critcal thinking skills that let him use these tools effectively. I am concerned that AI tools used non-critically will get users into a lot of trouble.
Recently I asked Perplexity to build me a chart of the halucination rates for all of the current models. It has improved, but the average is around 10%. 1 out of 10 answers provided by an LLM is just wrong. If one is using the tool to research something for which you are not an expert how will you be able to ascertain which answer is that 1 that will get you in trouble?
We also need to stop with this myth that anyone can build software applications. Sure, if the scope is small enough and the impact of bugs is small enough then have at it. For all of the real work, you still need a strong computer science background. Much like thinking, there are people who can do it and most who cannot.
We need to make sure our rational optimism stays rational.