"Although seemingly mysterious to the untutored, with sufficient data and knowledge, complicated systems can be understood with enough precision to be modified accurately and predictably"
But he is wrong about AI. Fundamentally we cannot steer AI systems due to their layers and we cannot get reliably get exact outcomes from them.
"Although seemingly mysterious to the untutored, with sufficient data and knowledge, complicated systems can be understood with enough precision to be modified accurately and predictably"
But he is wrong about AI. Fundamentally we cannot steer AI systems due to their layers and we cannot get reliably get exact outcomes from them.
I tend to agree with you. I had the opportunity to make a query to the Perplexity AI on Saturday. It made an answer which I copy pasted then before asking a follow up question Google ended the session. Later, I used Firefox privacy window and put up Start Page in it and went to Perplexity and asked the same question. It gave a similar but not quite the same answer. This time instead of copy paste, I made screenshots of everything the AI wrote. I was also able to get follow up questions. Perplexity put up sources at the top of the answers which I visited and verified how perplexity came to the answers it gave me. This reminded me of the "expert systems" that were being developed in the 1980s and 90s. What has really evolved is the ability to scan such huge parts of the internet to gather info at such rapid speed. The most amazing thing to me is the ability to give its answers in a straight forward conversational way, as if I was emailing a person who was an expert on the subject. But as with humans with credentials, I would not take what they say at face value. There, is where the sources it included were valuable to check. On of its sources I checked had more data that varied in some places that I would have to say the AI answer was not as valid as it would seem. Several other sources was a bit superficial. Assuming an AI is the last word on its answers will make people lazy and may cause expensive or dangerous errors when we apply them in real world settings such as medical decisions, or engineering a building or bridge.
"Although seemingly mysterious to the untutored, with sufficient data and knowledge, complicated systems can be understood with enough precision to be modified accurately and predictably"
But he is wrong about AI. Fundamentally we cannot steer AI systems due to their layers and we cannot get reliably get exact outcomes from them.
I tend to agree with you. I had the opportunity to make a query to the Perplexity AI on Saturday. It made an answer which I copy pasted then before asking a follow up question Google ended the session. Later, I used Firefox privacy window and put up Start Page in it and went to Perplexity and asked the same question. It gave a similar but not quite the same answer. This time instead of copy paste, I made screenshots of everything the AI wrote. I was also able to get follow up questions. Perplexity put up sources at the top of the answers which I visited and verified how perplexity came to the answers it gave me. This reminded me of the "expert systems" that were being developed in the 1980s and 90s. What has really evolved is the ability to scan such huge parts of the internet to gather info at such rapid speed. The most amazing thing to me is the ability to give its answers in a straight forward conversational way, as if I was emailing a person who was an expert on the subject. But as with humans with credentials, I would not take what they say at face value. There, is where the sources it included were valuable to check. On of its sources I checked had more data that varied in some places that I would have to say the AI answer was not as valid as it would seem. Several other sources was a bit superficial. Assuming an AI is the last word on its answers will make people lazy and may cause expensive or dangerous errors when we apply them in real world settings such as medical decisions, or engineering a building or bridge.
as someone said, "giving their brains over to the iphone."