Thinking Machines: The Quest for Artificial Intelligence - and Where It’s Taking Us Next - Luke Dormehl Audiobook

This post has 2 comments with rating of 5/52

• 
  BlueSun03
  November 5th, 2023
  

  Interesting. I plan to read it. I have been working as a programmer with AI since the late 1990s and keep up-to-date with current & potential future developments. Right now we have whats called generative artificial intelligence. Each application has to be trained & error-corrected for specific areas of functionality (like reading MRIs for example, where the best radiologists, shown any given MRI will read it with 85% accuracy while an AI program will score 95%). It takes humans to develop the datasets & then run thousands of batch “gradient descents” on sample neurons (called perceptors in AI talk). For each level of neurons we run thousands of gradient descents looking for the lowest “fail rate” an indication of accuracy). Imagine a 2-dimensional topographic map where we are picking random spots & trying to find our way to the lowest spot we can on the map. Then we “back propagate” to the inputs to each perceptors, each of which has a “weight” variable that indicates its estimated importance. We vary these input weights & repeat the gradient descents. When we thing we have this as accurate as possible (by comparing actual results to expected ones, we repeat with the next layer of the neural net until we reach the final output & check it against expected results to our training set’s data. This is no current threat because the AI is totally dependent on humans to provide the data & then train or fine-tune the AI’s neural network. It requires a huge amount of computing power & energy to do this. It took around a year and a half to train OpenAI’s ChatGPT 3.0 (they have gone through 3.5 & 4.0 but those cost $20.00 a month to have access to. Within a year or so, expect that we will move on to AGI (Artificial General Intelligence) where an AI app is flexible enough to learn in a wide varieties of domains - we are coming closer to awareness of agency, or self-awareness. When AGI is mated to Quantum Computers in the next few years, the process of training & tuning neural networks will be cheap & quick - maybe a million times faster. At that point, expect ASI (Artificial Super Intelligence) to emerge. This will happen quickly as it is the Quantum Computer that provides the power to make a million-fold jump in intelligence virtually instantly when they are mated. At this point, AI will be significantly, maybe vastly more intelligent than the human brain is able to achieve. At this point, sentience or awareness of self & agency is most likely to emerge. But, is it a threat? Will AI decide humans are a threat that needs to be wiped out? Or will the ASI system’s intelligence be so superior to ours that we pose no threat to it and it can merely let us continue on as we otherwise would - except for the incalculable value that intelligence of that magnitude can provide to human society - the ability to solve quickly virtually all of our current problems & crises & make our lives ones of leisure.

  BTW, as AI advances over the next several years, the need for human labor for the vast majority of work will disappear. We need to be working TODAY on new economic systems that no longer require humans to earn tokens (”money”) through labor to pay for our necessities & goodies. That means Capitalism must die and die soon or the loss of work will result in chaos & collapse. With AI doing all of our work, however, if we come up with a way to simply share the prosperity & productivity advances it makes, we could all create a total leisure society where, for the first time in history we are not chained to having to work to receive all the benefits of skyrocketing productivity (probably in the form of a Universal Basic Income greater than enough to meet all of our needs & wants). That particular question of chaos vs paradise is up to human intelligence to deal with and time is running out. Will we do the right thing? Will it be the failure of human intelligence, not the ascendency of artificial intelligence that really dooms us? I suspect, given what little regard I have for the intelligence of most humans I’ve met in my 75 years, that’s the real threat to our survival as a species.

  Thinking Machines: The Quest for Artificial Intelligence - and Where It's Taking Us Next - Luke Dormehl
• 
  byond_zen
  November 13th, 2023
  

  

  THANKS for sharing!

  Thinking Machines: The Quest for Artificial Intelligence - and Where It's Taking Us Next - Luke Dormehl

Add a comment