I love your subheading, it's so true re: telomeres :-) AI is such a double-edged sword, it's amazing to have a personal research assistant, less amazing to have one who lies some random percentage of the time, but disturbing to see AI already in use for mission-critical applications (what's the bar for letting AI diagnose me and dispense customized medicine? I hope it's sky-high!).
Thanks for the comment Stephanie! I continue to see the nonsense of Chat-GPT etal just a tribute to human arrogance. It is estimated 50% plus of all of our brainpower is committed to vision. Therefore touch, smell, taste and hearing are bit players in how we put things together. They are all important I imagine and when a model can layer them together the results will be impressive.
Knowing this means to me that a parlor trick only based upon a random slice of text on the internet seems absurd as a basis for anything mission critical as you describe.
I've been observing the ability of AI to identify mushrooms and fungi via apps. So far, for plants, the apps are not bad with decent specimens and mediocre for fungi. I'm yet to see an app which can identify the kinds of specimens that my group of human experts deals with.
Identifying plants and fungi isn't a concentrated task, but it is something which requires a huge amount of stored knowledge, something computers do better than humans. I'll continue to watch with interest.
Very interesting Melanie. While the analogy is imperfect, the steady improvement of AI to read radiology images seems like a similar field. I think the expression "data is the new oil" is apt. The pattern that seems to repeat in most domains is there is a critical amount of data that leads to cascading insights. That seems to be what pattern recognition and convergence to insight seems to need. Your comment captures one of the essential points I was trying to make. Tools like Chat-GPT seem to have collected enough data to model the world of language. There really are not that many words to choose from and it appears we only need a subset before we see how often we repeat ourselves!!! Plants and fungi will require a much larger physical data set as it is image data. We are awed by what Chat-GPT and Bard can do but other domains of our senses (sounds, images, etal) might require more data for the neural networks that evaluate for patterns to begin drawing conclusions and insights. While it does not get near the attention of the media, this is why Google AI-Fold is much more impressive as a result. Google has freely shared the results of their insights for drug and disease research for the 300M+ proteins of the world. Which of the plant and fungi applications do you recommend?
While just my opinion, I believe that insights result similarly to the way they do for us. If you show a young child different examples of fruits, they will become pattern recognizers eventually and might discern something they've never seen before as a likely fruit. If you only show them bananas and pineapples they might struggle. HOWEVER, if they get to SMELL them as well as see them the COMBINATION of those senses will more quickly converge to a durable pattern of what is a fruit. This is why only companies that have been indexing data in MANY DOMAINS for many years will have the library of content necessary to train these new models. Find the company behind a plant application that has an AVID userbase that SUBMITS content they see and they will have a great chance to succeed!
You're right, the apps need a lot of users, and they still need people to validate the answers. However the challenge is that many users aren't reliable identifiers (based on what I've seen on various facebook groups). So the AI may be being misled.
What if there was an upside to AI? Your article almost guarantees it, even for the welders who don't like it now. Ive lived through how we went from Texas Instrument calculators w red screens ( and belt clips) to where we are now with computers w amazing memory for $25O in just 40 years. We fight change. It's always coming. Great article Mark.
THANKS SO MUCH Paul. The gist of the post is that AI is going to be a big change like many before. I believe if we avoid adversarialism this revolution will improve lives in ways we cannot imagine. I don't know what the post was but I threw out an amazing statistic I think. We invented the transistor in 1947/48. We now produce more transistors IN A YEAR than all the leaves on all the trees on the whole planet. They transformed our lives. Those early red-screened calculators were memorable!!!
Thanks Anne. I LOVE the sentiment of take care of the busy work! I sort of agree with you about how the split will emerge. I also believe the big elements will be ROI. I recently had my dog in for an x-ray. I would imagine that even though being a radiologist is very robust training, machine-read x-rays will just become more and more prevailing because there is a bunch of money to be made if you can read most x-rays that way. I think x-rays are another example of where humans versus machine come into conflict. Lots of us work hard to train ourselves to be able to concentrate. Tasks that require continuous concentration will be hard to avoid shifting to machine functions mostly because humans at their core are a mixture of focused concentration and sensory interruption. One of my favorite stories about early applications of AI/ML was airplane autopilot. Getting a human to concentrate on a long flight to maintain a flight path is a lot to ask! The first autopilots emerged in the mid 1910s!
So interesting! Along similar lines I once did an interview with a guy who started a temp agency that placed people with autism and aspergers—apparently their powers of focus and concentration can be fantastic for jobs like proofreading and general quality control, far beyond the norm. But an AI NEVER loses focus!
As far as writing goes, all that AI is doing is replacing all the writers who wrote like robots in the first place!
That interview sounds wild and interesting! I would imagine as we've climbed the ladder of complexity there is more room for the autism/asperger thing and for some things quite good! I figure what makes great anythings (writers, etc) is perhaps the range of what they notice! I am not impressed by all the nonsense bots generate and respond to on platforms like Twitter. I just feel badly for people who spend a preposterous amount of time interacting with them. They've set their bar low and are probably missing a whole range of interesting and fun things!
Great point about no one cursing the robot welders. PS we need welders. When my daughter thought she wanted to be a doctor her biotech fiancé said, “I will put you out of business.”
haha CK! It's sort of the human condition I fear about the welders. The selfishness/survival part of all of us elevate our own self-importance I think. Funny about the doctor / biotech comment. Creative destruction -- I know a young newly minted radiologist -- a perfect example of business that is finally changing and now its in the acceleration phase. Recently had the dog x-rayed and even the vet had an electronic remote radiology practice. Crazy.
Oh there are dog specialists for everything. Derm vet, Neuro vet-- I have one for my dog since he has idiopathic epilepsy-- there are pet acupuncturists. The list is long.
You always give us something to think about, Mark. I'm not as optimistic as you about AI providing solutions to human-created problems like war and inequality. So far, I've seen students use it to help them with ideas for their essays--or writing those essays for them, after which they make minor corrections. I'm not sure what I think about that. My husband said he asked ChatGPT to tell him who he was and it claimed he'd written a book that he hadn't. Then he asked it to write the book for him and it came up with chapter headings that sounded credible, but on closer examination weren't. Maybe someday AI will do great things, but so far, it's an imperfect tool.
In one of my earlier posts, I shared my perspective about the possibilities of machine-directed thinking. Perhaps I am cynical in some way. I believe that AI is focused on the better half of the human brain. That is what I explored in the super long four part post. My sense is the big problems like war and inequality are ROOTED in our rear brains. Maybe the onset of analyzing the big problems in a different way can lead us to solutions.
I have some catching up to do on your posts. I don't have a problem with the length--it's just that you often pack so much interesting and thoughtful information into your newsletters that I want to give them the time and attention they deserve. I'll let you know what I think.
Very cool! I'm going to play Guinea pig with the sound therapy - after reading about it here, I searched around Google and learned that listening to something called "binaural beats" for 20 to 30 minutes daily may help reduce anxiety. There are binaural beat tracks all over YouTube. I'll give it a try and let you know how it works our.
Maybe someone will write a modern John Henry story about AI....
Good for you Jeannine! I would expect, like most innovations there will be a non-validated claims for this sort of thing. I hope this works for you. I am a boring, evidence-based guy. This old post has the details of the medical evaluated examples off light and sound treatments. Alzheimer's is the big target. https://markdolan.substack.com/p/cleanup-in-aisle-3. The long-term AI/ML vendors have been at this since 2005 (Google DeepMind for example). I expect the MULTI-SENSE stuff will make its way into the mainstream in the very near future. An an animal lover I think we will see budding capacity to interpret pet language for example. It will start with whales and dolphins but I imagine the big money chase will be for understanding our pets. Thanks for reminding me of the John Henry story.
I posted about it recently. AI has begun to reveal sense couplings we never considered. Both sound n light appear to reverse plaque in the brain. Not new-age crystal stuff. They appear to trigger the same cleanup cycle properly sleep patterns do as brain maintenance. Akin to rebooting windows for memory leaks
I love your subheading, it's so true re: telomeres :-) AI is such a double-edged sword, it's amazing to have a personal research assistant, less amazing to have one who lies some random percentage of the time, but disturbing to see AI already in use for mission-critical applications (what's the bar for letting AI diagnose me and dispense customized medicine? I hope it's sky-high!).
Thanks for the comment Stephanie! I continue to see the nonsense of Chat-GPT etal just a tribute to human arrogance. It is estimated 50% plus of all of our brainpower is committed to vision. Therefore touch, smell, taste and hearing are bit players in how we put things together. They are all important I imagine and when a model can layer them together the results will be impressive.
Knowing this means to me that a parlor trick only based upon a random slice of text on the internet seems absurd as a basis for anything mission critical as you describe.
I've been observing the ability of AI to identify mushrooms and fungi via apps. So far, for plants, the apps are not bad with decent specimens and mediocre for fungi. I'm yet to see an app which can identify the kinds of specimens that my group of human experts deals with.
Identifying plants and fungi isn't a concentrated task, but it is something which requires a huge amount of stored knowledge, something computers do better than humans. I'll continue to watch with interest.
Very interesting Melanie. While the analogy is imperfect, the steady improvement of AI to read radiology images seems like a similar field. I think the expression "data is the new oil" is apt. The pattern that seems to repeat in most domains is there is a critical amount of data that leads to cascading insights. That seems to be what pattern recognition and convergence to insight seems to need. Your comment captures one of the essential points I was trying to make. Tools like Chat-GPT seem to have collected enough data to model the world of language. There really are not that many words to choose from and it appears we only need a subset before we see how often we repeat ourselves!!! Plants and fungi will require a much larger physical data set as it is image data. We are awed by what Chat-GPT and Bard can do but other domains of our senses (sounds, images, etal) might require more data for the neural networks that evaluate for patterns to begin drawing conclusions and insights. While it does not get near the attention of the media, this is why Google AI-Fold is much more impressive as a result. Google has freely shared the results of their insights for drug and disease research for the 300M+ proteins of the world. Which of the plant and fungi applications do you recommend?
While just my opinion, I believe that insights result similarly to the way they do for us. If you show a young child different examples of fruits, they will become pattern recognizers eventually and might discern something they've never seen before as a likely fruit. If you only show them bananas and pineapples they might struggle. HOWEVER, if they get to SMELL them as well as see them the COMBINATION of those senses will more quickly converge to a durable pattern of what is a fruit. This is why only companies that have been indexing data in MANY DOMAINS for many years will have the library of content necessary to train these new models. Find the company behind a plant application that has an AVID userbase that SUBMITS content they see and they will have a great chance to succeed!
You're right, the apps need a lot of users, and they still need people to validate the answers. However the challenge is that many users aren't reliable identifiers (based on what I've seen on various facebook groups). So the AI may be being misled.
We will always need quality training data for ML.
What if there was an upside to AI? Your article almost guarantees it, even for the welders who don't like it now. Ive lived through how we went from Texas Instrument calculators w red screens ( and belt clips) to where we are now with computers w amazing memory for $25O in just 40 years. We fight change. It's always coming. Great article Mark.
THANKS SO MUCH Paul. The gist of the post is that AI is going to be a big change like many before. I believe if we avoid adversarialism this revolution will improve lives in ways we cannot imagine. I don't know what the post was but I threw out an amazing statistic I think. We invented the transistor in 1947/48. We now produce more transistors IN A YEAR than all the leaves on all the trees on the whole planet. They transformed our lives. Those early red-screened calculators were memorable!!!
I like to think that the more AI takes over the busy work the more time we will have to employ the skills and talents that make us essentially human.
Thanks Anne. I LOVE the sentiment of take care of the busy work! I sort of agree with you about how the split will emerge. I also believe the big elements will be ROI. I recently had my dog in for an x-ray. I would imagine that even though being a radiologist is very robust training, machine-read x-rays will just become more and more prevailing because there is a bunch of money to be made if you can read most x-rays that way. I think x-rays are another example of where humans versus machine come into conflict. Lots of us work hard to train ourselves to be able to concentrate. Tasks that require continuous concentration will be hard to avoid shifting to machine functions mostly because humans at their core are a mixture of focused concentration and sensory interruption. One of my favorite stories about early applications of AI/ML was airplane autopilot. Getting a human to concentrate on a long flight to maintain a flight path is a lot to ask! The first autopilots emerged in the mid 1910s!
So interesting! Along similar lines I once did an interview with a guy who started a temp agency that placed people with autism and aspergers—apparently their powers of focus and concentration can be fantastic for jobs like proofreading and general quality control, far beyond the norm. But an AI NEVER loses focus!
As far as writing goes, all that AI is doing is replacing all the writers who wrote like robots in the first place!
That interview sounds wild and interesting! I would imagine as we've climbed the ladder of complexity there is more room for the autism/asperger thing and for some things quite good! I figure what makes great anythings (writers, etc) is perhaps the range of what they notice! I am not impressed by all the nonsense bots generate and respond to on platforms like Twitter. I just feel badly for people who spend a preposterous amount of time interacting with them. They've set their bar low and are probably missing a whole range of interesting and fun things!
Great point about no one cursing the robot welders. PS we need welders. When my daughter thought she wanted to be a doctor her biotech fiancé said, “I will put you out of business.”
haha CK! It's sort of the human condition I fear about the welders. The selfishness/survival part of all of us elevate our own self-importance I think. Funny about the doctor / biotech comment. Creative destruction -- I know a young newly minted radiologist -- a perfect example of business that is finally changing and now its in the acceleration phase. Recently had the dog x-rayed and even the vet had an electronic remote radiology practice. Crazy.
Oh there are dog specialists for everything. Derm vet, Neuro vet-- I have one for my dog since he has idiopathic epilepsy-- there are pet acupuncturists. The list is long.
Specialization has crept into everything. So many of us love our dogs.
You always give us something to think about, Mark. I'm not as optimistic as you about AI providing solutions to human-created problems like war and inequality. So far, I've seen students use it to help them with ideas for their essays--or writing those essays for them, after which they make minor corrections. I'm not sure what I think about that. My husband said he asked ChatGPT to tell him who he was and it claimed he'd written a book that he hadn't. Then he asked it to write the book for him and it came up with chapter headings that sounded credible, but on closer examination weren't. Maybe someday AI will do great things, but so far, it's an imperfect tool.
In one of my earlier posts, I shared my perspective about the possibilities of machine-directed thinking. Perhaps I am cynical in some way. I believe that AI is focused on the better half of the human brain. That is what I explored in the super long four part post. My sense is the big problems like war and inequality are ROOTED in our rear brains. Maybe the onset of analyzing the big problems in a different way can lead us to solutions.
Obviously I should read those posts!
They are probably too long. I may try to rewrite them someday. I think the story is great but too long. https://markdolan.substack.com/p/how-to-tame-your-lizard-part-1
I have some catching up to do on your posts. I don't have a problem with the length--it's just that you often pack so much interesting and thoughtful information into your newsletters that I want to give them the time and attention they deserve. I'll let you know what I think.
THANKS!!! Part 3 has a fictitious person going into Neural Depot to buy a brain -- I was particularly proud of that part of the post :)
Very cool! I'm going to play Guinea pig with the sound therapy - after reading about it here, I searched around Google and learned that listening to something called "binaural beats" for 20 to 30 minutes daily may help reduce anxiety. There are binaural beat tracks all over YouTube. I'll give it a try and let you know how it works our.
Maybe someone will write a modern John Henry story about AI....
Good for you Jeannine! I would expect, like most innovations there will be a non-validated claims for this sort of thing. I hope this works for you. I am a boring, evidence-based guy. This old post has the details of the medical evaluated examples off light and sound treatments. Alzheimer's is the big target. https://markdolan.substack.com/p/cleanup-in-aisle-3. The long-term AI/ML vendors have been at this since 2005 (Google DeepMind for example). I expect the MULTI-SENSE stuff will make its way into the mainstream in the very near future. An an animal lover I think we will see budding capacity to interpret pet language for example. It will start with whales and dolphins but I imagine the big money chase will be for understanding our pets. Thanks for reminding me of the John Henry story.
I didn't know that about sound therapy. I've only just heard of it from a massage therapist in town. What a world!
I posted about it recently. AI has begun to reveal sense couplings we never considered. Both sound n light appear to reverse plaque in the brain. Not new-age crystal stuff. They appear to trigger the same cleanup cycle properly sleep patterns do as brain maintenance. Akin to rebooting windows for memory leaks
I must have missed that one. That’s fascinating. Maybe I’ll try out the offering that’s come up locally!
The optional links and footnotes in https://markdolan.substack.com/p/cleanup-in-aisle-3 are quite good I think.