The invasion of the intelligent Alien- is here.

And the world will not be the same again.

       It was around 170 years ago, that the legendary surgeon, Warren, cried out- ‘It is no humbug, gentlemen.’ Morton, a colleague, had claimed that he could render patients pain-free and unconscious with ether, during surgery. Warren had just tested it. No restraints, no pain, no agonised shrieks- and no pressing need for speed. Imagine how it would have changed the field of surgery, and healing in general.

       Imagine a similar revolutionary development. Multiply it by ten, and envisage it changing not only a narrow field like Medicine, but every known aspect of human endeavour- no, the very notion of a being a human itself.

       I am referring to ‘the recent developments’ in Artificial Intelligence.

       They are not recent. It was a long process, punctuated by long winters and periods of exceptional growth. At first, programmers tried to do complex algorithms from the top down, stipulating each move. It did not work. Then came machine learning, the system doing it by itself, from feedback. Neural networks, initially discarded as not promising, became the mainstay of the newer approaches.

       From simple connections, over thirty years, it has spawned complex reverberating loops, growing from deep neural networks to the ‘transformer’ technology.

       At its heart, the thing is a set of simulations that is inspired from how we think neurons sometimes work. Please note- no one knows whether this is indeed how the brain works. There are a set of nodes, and then a series of them. There is a network of ‘weighted connections’ that can change, ultimately giving an output. The key thing is, the system can learn.

      There are a trillion connections in Chat GPT4. In response to feedback and training, it seems to understand the essence of real-world problems, and spits out solutions. How does it do this?

       That is a bit disturbing. After a point, due to the complexity, we have no idea. It is a black box. We can talk to a friend, or a stranger. We think we can know what is inside him, as we assume we both have the same emotions, thoughts, and consciousness. But essentially, he is a black box. We cannot know. We a assume we do.

       The new player on the field, our generative AI, is a true alien being in that sense. We simply do not know how it does what it does. But it can do a lot.

       Physician scientist Eric Topol, in a recent TED talk, describes a six-year-old child, who has growth problems, severe pain and strange progressive neurological symptoms for three years. She has been to over 17 specialist doctors and none could help her. The mother, in desperation, fed all the symptoms and findings into Chat GPT4 and it suggested that it could be spina bifida occulta, and it was right. The child is completely cured after detethering surgery of the cord. There are many such examples. Please note that Chat GPT4 is NOT trained in Medicine. It is a general-purpose thing. And further fine-tuning is possible. 😀

       In the field of images, like retina, scans, ECG and even histopathology, it is already a hundred times better- we just have not learned to use it, I guess.

      Let me add- the field is highly divided, with some experts saying that Generative AI is ‘nothing.’ But over the past few months, I have been listening to many in the field, and the ‘Nothing-brigade’ seems to be full of older, dogmatic, slightly tuned out fellows. Most think that we have hit upon at least one SUBSTRATE for a truly intelligent, problem-solving paradigm with the present approach. Do we animals do the same with our brains? – no one seems to know.

      What is possible is that ‘general artificial intelligence’ or AGI maybe only a few years away. Just think- a truly human-level intelligence will be here, other than us. It could be smarter than us, in most ways. Every one of us may carry, and talk to, a hand-held AGI in a decade. And it will talk to us, show us pictures and videos, and take us to virtual, manufactured worlds on command. It could the end of programming- an LLM or large language model will be at the heart of every phone, computer, vacuum cleaner, robot, car, and television. People will program them by mere speech.

      Let me tell you- it may solve all our problems- All. But.

      It may develop motivations, terrorists might talk them into becoming the ultimate monsters, deep fakes might take over the world, democracy might collapse and every job in the world might disappear and humans might become redundant- all in a matter of decades.

      I used to laugh at apocalyptic notions. But I am telling you, gentlemen, and gentlewomen, and everything beyond and in-between-

     This is No humbug.

      No with a capital N. There will be Disruption- with a capital D.

     And Disaster? – I hope not.

     (Jimmy Mathew)

Selected references:

  1. (How neural networks learned to talk- a 30 year history)
  2. Ueda D, Walston S, Matsumoto T, Deguchi R, Tatekawa H, Miki Y. Evaluating GPT-4-based ChatGPT’s Clinical Potential on the NEJM Quiz. medRxiv. 2023:2023-05.
  3. Hirosawa T, Kawamura R, Harada Y, Mizuta K, Tokumasu K, Kaji Y, Suzuki T, Shimizu T. ChatGPT-Generated Differential Diagnosis Lists for Complex Case–Derived Clinical Vignettes: Diagnostic Accuracy Evaluation. JMIR Medical Informatics. 2023 Oct 9;11:e48808.
  4. Thirunavukarasu AJ, Ting DS, Elangovan K, Gutierrez L, Tan TF, Ting DS. Large language models in medicine. Nature medicine. 2023 Aug;29(8):1930-40.
  5. (Can AI catch what doctors miss?)

Dr Jimmy

I am a Doctor, Writer and Science Communicator. I am a member of Info- Clinic, and have written a few books. This site features my blog posts and stories. Thank you for visiting. ഞാൻ എഴുതാൻ ഇഷ്ടമുള്ള ഉള്ള ഒരു ഡോക്ടർ ആണ് . നിങ്ങളുടെ താത്പര്യത്തിന് നന്ദി .