Will ChatGPT help longevity breakthroughs?
Вставка
- Опубліковано 10 жов 2024
- You have most likely heard about ChatGPT - the new AI chatbot tool from OpenAI . [ChatGPT](chat.openai.co..., which was released at the end of last year, is trained on OpenAI’s family of large language models and creates surprisingly intelligent-sounding text in response to user prompts. More than one million people tried it out in the first week - including me - and me being the normal person that i am and took full advantage of it,
Now, besides entertaining myself while procrastinating my thesis, or for doing basic info searches or writing science essays - I was keen to know, as you may be too, will ChatGPT actually help us make science discoveries faster in particular in biological research such as coming up with drug solutions to cancer, aging, alzheimers?
Find me on Twitter - / eleanorsheekey
Support the channel
through PayPal - paypal.me/shee...
through Patreon - / thesheekeyscienceshow
Please note that The Sheekey Science Show is distinct from Eleanor Sheekey's teaching and research roles at the University of Cambridge. The information provided in this show is not medical advice, nor should it be taken or applied as a replacement for medical advice. The Sheekey Science Show and guests assume no liability for the application of the information discussed.
Icons in intro; "www.freepik.co..."Background vector created by freepik - www.freepik.com
Have you tried ChatGPT - what other ways do you think it can help science?
It helped me restore NAD metabolism and mitochondrial quality control without expensive Fads like NMN that the FDA is trying to relegate to only be available for the rich.
Jokes on them, its cheap and easy.
Pity Stratoai ChatGPT search engine shut down. I smashed through 6 months research in days.
I had a bunch of questions lined up to help break through a couple more mysteries, but was locked out :(.
Oh well, at least I worked out half of what I wanted. Why nucleotide synthesis becomes dysfunctional with age, and how adenine and guanine don't really cause gout. Mitochondrial dysfunction and lack of nitrated CLA, nitrated fatty acids, or nitro-Oleic Acids is what causes it.
Probably linked to NO2 metabolism dysfunction and arginase dysfunction. Nitrogen is an important base for Nucleotides, so hardly surprising there is more stemming from those pathways. Might already know how to fix it. (need to test)
AKA too much processing of Oil Fractions from Seeds, Grains, and Nuts and Milks. Need to leave them unaltered. Even then, you probably won't get enough.
I'm becoming more and more aware that a lot that drives aging stems from diet.
over-rated, mostly entertaining, just laughs
It'd help if they had an uncensored version that didn't seem to kneecap any questions about medicine or politics.
I believe this tech will be immensely useful, but, like you said, it is a tool. It can be great at aggregating large quantities of information and communicate the results with us in a natural manner, but in its current state, I wouldn't ask it to draw conclusions based on that information. That may change over time, though.
"But I would rather chatgpt be honest upfront and tell me it did not know the answer, than to give me a response that I know to be wrong" => this is because it does not actually "know" anything, all it does is select the most likely next word in a sequence of words. It might look smart, but that is only because it has been trained on an inconceivable amount of data for months. If it is wrong though, you can point out to chatgpt that it is wrong, and sometimes it will fix the answer. But it might also give you a different answer that is just as wrong as the first one and be conviced that this is the correct answer. That being said, I finished my master's degree in CS/AI in 2017 and back then I thought language models that are as good as chatgpt were at least a decade away. The speed at which things are moving now is both exciting and scary at the same time.
This is interesting. I already did things like this too. Very useful when driven properly
Happy new year! This is a wonderful channel.
I have fooled around with ChatGPT a little, mostly trying to get it to explain physics concepts that I'm fuzzy on. It helped a little, but not much. As you said, the information was very "generic." (Great description, btw.) It tended to describe things in a dictionary sort of way, and when I asked it to expand on certain points it tended to use the same words and phrases in slightly different order. A human teacher will understand where a student is confused from the nature of their questions, and know how to come at the subject from a different angle to make it more comprehensible. ChatGPT doesn't understand anything, of course. So like you said, it's up to the user to identify their own areas of ignorance and try to find questions with that different angle that can unlock the answers from the AI.
I'm sure all those essays and papers I had to write in college and law school took years off my life. So, yes, it will help longevity.
😂
I think Mogan Levine identified it best that aging will be fixed through expotential growth of AI and Increasing Processing Power of Computers. She sounded very positive about the next 10- 30 years.
ChatGPT isn't an AI. It's a Language Model. AlphaFold is an example of a REAL AI . The combination of ChatGPT and real AI like AlphaFold is going to be to be a new leap in AI adoption and usage / development. It's the REAL AI models advancement that's going to help eventually solve aging, But ChatGPT can be a significant catalyst towards the adoption, investment, research, and development needed to make that advancement happen.
You can ask chatgpt to list the sources it found the information
GPT 4 has made substantial improvements. Also, I would really love to hear what you think of other AI tools being used in research like Elicit AI. Please make more videos exploring other research AI platforms!
I think chat GPT was updated recently so it now is up to date through mid-Dec 2022. Originally, it was trained on 2021 data only.
Great video. I encountered the same thing when using it for debugging software package interdependencies.
Just this morning I was playing with it on more philosophical topic. It has absolutely no problem giving answers that are grammatically perfect, but self-inconsistent.
I think it's excellent at giving a correctly shaped answer, but the contents of the shape are highly suspect.
Or: Finger(chat gpt) -> Moon(actual answer somewhere out in the world)
We need an update from a scientist in the field that is into science communication, how have tools like GPT 4.0, Gemini Pro (soon ultra ;) ), Perplexity AI, consensus etc. impacted you and your research and how will it impact scientific discovery in the years to come. How can big data and machine learning aid halt or slow down the complex and intricate process of ageing?
This is very interesting..I share your approach to using it and not treating it as gospel but equally not dismissing it as a useless toy as some are keen to do. Think its great the way you can build questions upon questions and some ways find additional insights or clarity to what you're searching for. The code generator and the general way it can framework mundane tasks is pretty helpful. I've also seen something about people using Dall-E the image generator to generate images of protein structures based on text prompts. Really nifty!! Its a pretty exciting thing for sure. I hope someone as bright and inquisitive as yourself will play a part in using the bones of Chat GPT and training it up on specialist knowledge and data that you are focused on. It sounds like specialisation for it is where we're about to go next..have a crack at it, you could invent the future! Peace from Ireland!
ChatGPT is too “generalized” and needs to be trained on a specific data set to be of value to any one field. For example, PythonGPT might only do python coding. Plus these models are trained and that takes time so the information is immediately stale upon release. That said, this type of model could apply to any research area, industry, or even a company. This type of specific training opens up job opportunities. Also the term “prompt engineer” has been tossed around. One new video referred to a prompt engineer as an “AI Whisperer.” Can’t begin to imagine what Chat GPT v4 will unleash. Gonna be a wild ride!
If you ask chatgpt for references, it does give some of them to you, although the references often contradict what chatgpt is saying, its in the right ballpark.
I like ChatGPT. But its effectiveness is still based on how one crafts the context and clarity of the question.
Human critical thinking is still required for generating new knowledge when it comes to machine learning AI systems. And I've come to understand that the real game changer comes from integrating along with the Chrome extensions ( Merlin, WebChat GPT) which helps overcome a number of limitations like expanding the knowledge base past 2021 and integrating with other websites and common apps.
It is possible to ask it for help to set up a scienifically sound experiment.
ChatGPT suffers from the same thing we do. I doesn't know what it doesn't until it gets it wrong. Perhaps they need to add a "meta heuristic"? Explanation: Code to allow ChatGPT to learn how and how much it has learn. I know this might be a tall order.
The problem is it doesn't "know" anything, as it told me. ChatGPT just understands the patterns of human language. It's incredibly good at predicting each word that comes next in a sentence, but it's unaware of anything that may be missing or factually wrong in the dataset that it was trained on.
@@Kevin_Street ele já se tornou mais inteligente do que você Kevin
I realize that this topic is outside of your scope of practice/knowledge areas. However, you might benefit a large number of your viewers (who are laypeople but are nonetheless interested in helping to advance biogerontology research via making donations or investing in start-ups), if you could create a video that looked at the funding landscape for biogerontology-related projects. You could ask and answer questions like, "Is the field under-funded?"/"If the field is underfunded, how much of a shortfall exists?"/Where are the shortfalls the most severe?/etc.
It's important to recall ChatGPT is a LANGUAGE MODEL, NOT an "Intelligence" . It has intelligence in it's LANGUAGE ability, but does NOT have logical inference, Deductive reasoning capabilities, or anything along those lines. In the same way that an average Human American Adult can be very good at the English Language, While lacking the slightest ounce of actual intelligence : Fox News, For Example. As well, This isn't "new" AI technology even, We've had online AI Chatbots for well over a decade now. However ChatCPT DOES seem to be a revolution in language Ability, in terms of being able to create human like responses and provide information in a back and forth manner. It SEEMS like intelligence, But it actually isn't so, and therefore won't "Help", per se, in longevity research more than any other tool, as mentioned.
AlphaFold, On the other hand, IS a big deal, as It's an AI Model that's solved a previously unsolved problem of predicting Protein Structures with high accuracy -- Something that IS "Intelligent" able to provide humans with New Data. AlphaFold can answer questions for humans that Humans do not know, Which ChatGPT Cannot do. AlphaFold can give Researchers NEW information that they can use to form new ideas and theories, Whereas ChatGPT can only talk about things it's already been made explicitly aware of by other humans (The internet).
The Power of ChatGPT will be COMBINING the Practical Language Capabilities WITH the Intellectual Power of AlphaFold , And Any limitless number of other similarly powerful technologies. When that happens, You'll be able to ask ChatCPT something, which will interpret the intent and communicate that interpretation to AlphaFold's Capabilities to DISCOVER the answer, and finally AlphaFold sends the answer to ChatGPT, Who can then interpret the data and output a Human-Level Language answer as a response.
^ When these sorts of integrations start happening, That's when cool stuff will start happening. ChatGPT's current ability to code is a very dull psudo-version of this. It's just Scraping code segments from various places around the web and putting them together just like we actual software engineers do. It's ability to make functional code is PROBABLY related to it's ability to speak English properly , The two are incredibly similar.
Bonus Use : If you're running a large X plant or Y Factory or Z Network, For example, an integration with ChatGPT could be used to execute physical commands and diagnose situations and understand/resolve potential issues, Using Natural Language as opposed to Computer Code or Physical Buttons / Monitors.
AI will be a great tool for science and medicine.
Looks like chat gpt won’t be replacing you anytime soon. Ur safe 🤣🤣🤣
For now 😂
Lifespan. Healthspan. Fertilespan.
As health span increases, the retirement age should increase too.
People should have time and money for a second specialization.
Dementia, unemployment, illiteracy, and infertility should decrease.
As the percentage of supercentenarians increases, who remains fertile?
Our nuclei slowly accumulate mutations, till nobody remains fertile.
The field of genome editing has a big future.
It could solve even the problem of mutagenetic insertion.
It could repair and conserve genomes and epigenomes.
Otherwise reproduction cannot be forbidden or diminished forever.
Anyone should have the right to die.
But nobody should have the duty to die.
We should never get so far as to see manifestations for the death of some generations.
There are many folk or pragmatic, traditionalist politicians and manipulators who think that geroscience and transhumanism have nothing to do with genuine politics.
Illusory, strange, unrealistic topics, movements.
There are no money and frameworks, conditions for such madness.
These are just words posted in vain...
Paradise lost.
Nothing but Schengen really matters for them.
Society is completely shocked and unprepared for such madness.
most of the sciences fail already at 5
I think you should show this video to the people at OpenAI. It may sound weird to do that, but this kind of specific feedback from a highly educated person is much more valuable to them than the random feedback they usually get, trust me. Please send them an email or something.