ChatGPT test at LENR-CANR.org

  • I am "hyping" it only because it is impressive.

    As I said: For private use as a summary writer its OK as you have full control over the knowledge base. But its public use is already very damaging. It's use for Orwell truth ministry like "fact checking" that in reality is marketing fake news spreading is widespread.


    Most money with GPT is made in private porn talks...

  • As I said: For private use as a summary writer its OK as you have full control over the knowledge base. But its public use is already very damaging. It's use for Orwell truth ministry like "fact checking" that in reality is marketing fake news spreading is widespread.


    Most money with GPT is made in private porn talks...

    Autonomous GPT Agents is even more profound. Are you familiar with Python?
    It's pretty incredible what one can create for research purposes and writing content.

  • ChatGPT is a very good tool to learn about common knowledge to a very high level. Now if you want to brake out of a bubble it's pretty useless. Also when it comes to technical fields, for computer science and programming it is a very useful tool. I wish I could use it more, but we can't share our code, so I use it to explain API and how to perform tasks. Perhaps 5-10 tasks a day is helped by ChatGPT to solve problems in minutes which would may times take hours or days, that's hours a day saved and the trace of that I don not care. Also it's much better at doing calculations now than before, the nice thing with ChatGPT is that you get the trace and can validate the calculation and hence as a human I feel in control of what I'm using. I also never copy and paste without knowing the solution is correct.

  • ChatGPT is a very good tool to learn about common knowledge to a very high level.

    I consider it a superb indexing system. Far better than conventional SQL.


    Now if you want to brake out of a bubble it's pretty useless.

    That depends on what data is loaded into it. The dataset in the public version of ChatGPT appears to be mostly conventional sources. The dataset at the the ChatGPT running at LENR-CANR.org is every paper in the Library. So, if you are scientist who knows little about cold fusion other than what Nature or Wikipedia say, you can break out of that conventional bubble by asking the LENR-CANR.org ChatGPT. She will tell you about every mainstream paper, and also about papers that even cold fusion researchers consider beyond the pale. Ask her this:


    Do you have papers on alchemy?

    Give her 30 seconds or so to think about it. Here is a typical response:

    Yes, there are several papers available on alchemy. One paper is "Alchemy: Myth or Science? Rejoinder to 'The Curious Case of an Experiment with Alchemy'" by M. Srinivasan, which discusses the history and scientific basis of alchemy. Another paper is "Search for Isotopic Anomalies in Alchemical Silver Coins from the Germanischen National Museum in Nuremberg" by H. Bottollier-Curtet et al., which analyzes six silver coins of supposedly alchemical origin and concludes that the silver in these coins is most likely of natural origin. There are also papers on the topic of biological transmutations, which were studied by some alchemists, such as "Biological Transmutation. Natural Alchemy" by L. Kervran and G. Oshawa.


    (You can ask it for the URL for these titles. Try this: What is the URL for the paper "Search for Isotopic Anomalies in Alchemical Silver Coins . . ."?)


    The ability to load data limited to one website will make Chatbots that break out of conventional beliefs. I expect the Flat Earth Society will soon have one.

  • Are you familiar with Python?
    It's pretty incredible what one can create for research purposes and writing content.

    I am not familiar with Python. I could figure it out, but it would take time. (That's my excuse, and I am sticking with it.) If you could help me set up Python to work with the Natural Language Tool Kit (https://www.nltk.org/) on my computer, I would appreciate it.


    I might be able to use NLTK to improve the raw output from the Amazon AWS Textract OCR program that I use for old documents, such as the ICCF3 book. I have written a Pascal program that does a pretty good job, but I gather NLTK might work better. I need to break the text into paragraphs. ChatBots need to have text in short paragraphs. A paragraph that exceeds 512 tokens confuses the poor things. They lose track.


    If it could eliminate page headings, that would be even better.

  • My comment was on the public version of ChatGPT. But yes with the right training you would have some power, but we will not get that for free. I also wonder how many scientifc articles really they have fed the system and I suspect that there are som martingales added to it so that crazy people suddenly will get receipts to do all kinds of wild things like toxic gas and what not.

  • My comment was on the public version of ChatGPT. But yes with the right training you would have some power, but we will not get that for free.

    It is not free. It is costing me $150 a month.


    I also wonder how many scientifc articles really they have fed the system and I suspect that there are som martingales added to it so that crazy people suddenly will get receipts to do all kinds of wild things like toxic gas and what not.

    For the LENR-CANR.org bot, I know exactly which scientific articles were fed into the system, because I fed them myself. They include every paper in the Acrobat folder:


    Index of /acrobat


    However, some of them do not show up often in Bot searches because they are in old versions of Acrobat format. One of the ones in Japanese was reduced to a bunch of numbers and punctuation marks. It wasn't working at all. I asked the vendor, who looked into the data file and found only this:


    2004

    5

    22

    2.45eV

    1989 6 3

    ( ) 10cm 1cm

    1989 1948 23 1 ( )

    ( ) ( ) ( ) ( ) 30 1

    ( ) 1971 1979 2 Texas A&M Texas (1980 1982 ) Bockris 2 2 1000 Haupin 10 Haupin 1

    . . .


    I have to convert all files to text, with paragraphs.

  • Just on a note of curiosity JedRothwell , have you noticed increased traffic to LENR-CANR.org since you made the chatboxes available?

    I do not see a dramatic uptick. It hasn't been there long . . . Monthly totals are in a spreadsheet and graph here:


    LENR-CANR.org Total Downloads


    This month it is running at 253 per day (7,847 per month).


    I would like to know how many queries the Bot is getting. I may not be able to find out unless I turn on the feature to have readers to register before they ask a question. They would never do that. That would also give me a copy of reader queries, I think.


    The ChatFast bots are intended for things like customer support. For that, you might want people to register, and then have the queries copied to a support person. You can also input a list of Frequently Asked Questions and canned answers. I think the bot recognizes variations in the questions. See:


    ChatFast - Create GPT chatbot from your own data

  • Nice! I am guilty of about 5 of those, give or take.

    I certainly Hope to see LENR helping humans to blossom, and I'm here to help it happen.

  • Carnegie Mellon University is running an AI supported catalyst discovery project. The website reveals something of their methods.


    Open Catalyst Project
    Using AI to model and discover new catalysts to address the energy challenges posed by climate change.
    opencatalystproject.org


    The Open Catalyst Project is a collaborative research effort between Fundamental AI Research (FAIR) at Meta AI and Carnegie Mellon University's (CMU) Department of Chemical Engineering. The aim is to use AI to model and discover new catalysts for use in renewable energy storage to help in addressing climate change.

    Scalable and cost-effective solutions to renewable energy storage are essential to addressing the world's rising energy needs while reducing climate change. As we increase our reliance on renewable energy sources such as wind and solar, which produce intermittent power, storage is needed to transfer power from times of peak generation to peak demand. This may require the storage of power for hours, days, or months. One solution that offers the potential of scaling to nation-sized grids is the conversion of renewable energy to other fuels, such as hydrogen. To be widely adopted, this process requires cost-effective solutions to running chemical reactions.

    An open challenge is finding low-cost catalysts to drive these reactions at high rates. Through the use of quantum mechanical simulations (density functional theory), new catalyst structures can be tested and evaluated. Unfortunately, the high computational cost of these simulations limits the number of structures that may be tested. The use of AI or machine learning may provide a method to efficiently approximate these calculations, leading to new approaches in finding effective catalysts.

    To enable the broader research community to participate in this important project, we have released the Open Catalyst 2020 (OC20) and 2022 (OC22) datasets for training ML models. These datasets altogether contain 1.3 million molecular relaxations with results from over 260 million DFT calculation

  • ChatGPT is a very good tool to learn about common knowledge to a very high level. Now if you want to brake out of a bubble it's pretty useless. Also when it comes to technical fields, for computer science and programming it is a very useful tool. I wish I could use it more, but we can't share our code, so I use it to explain API and how to perform tasks. Perhaps 5-10 tasks a day is helped by ChatGPT to solve problems in minutes which would may times take hours or days, that's hours a day saved and the trace of that I don not care. Also it's much better at doing calculations now than before, the nice thing with ChatGPT is that you get the trace and can validate the calculation and hence as a human I feel in control of what I'm using. I also never copy and paste without knowing the solution is correct.

    All my work is open source, it's just a matter of making the time to publish it.

    I will post updated codes to my Github with instructions.
    Make no mistake, it's costing me money for OpenAI and Serapi, but I am getting a good extended intelligence along the way and an ability to :)

    I have no chance competing with the amount of information they have, but if I can interface with it... that would be something really special. ;)

    Density Function Theory will help expand my models general

  • There are alway unintended consequences. The use of AI in music has led to the strange situation of fake listeners streaming fake music...


    Spotify Purges Thousands of AI Songs to Stop Bots From Streaming Them
    In addition to fake music, artificial intelligence has created a big new problem for Spotify: fake listeners. It's an infinite loop of "artificial streaming."
    gizmodo.com

    "The most misleading assumptions are the ones you don't even know you're making" - Douglas Adams

  • All my work is open source, it's just a matter of making the time to publish it.

    I will post updated codes to my Github with instructions.
    Make no mistake, it's costing me money for OpenAI and Serapi, but I am getting a good extended intelligence along the way and an ability to :)

    I have no chance competing with the amount of information they have, but if I can interface with it... that would be something really special. ;)

    Density Function Theory will help expand my models general

    FYI I'm also an open source developer and I spent much time with that when I'm not poking a little with alternative theory. I have an ordinary day job as a programmer. Right now I cannot work as they temporarily closed down all the servers that we used. Anyway the freedom to do do what I like is more important to me than making a ton of money or chase grants to do research that stay in line with those in power.

  • There are alway unintended consequences. The use of AI in music has led to the strange situation of fake listeners streaming fake music...


    https://gizmodo.com/spotify-ai…istening-songs-1850419032

    I have been suggesting a fix for this for a couple of years now. AI already has tokenization for there monetary rails. A good solution for ensuring privacy and authenticity with the advancement of AI is tokenization on blockchain or other encryption networks. People can than verify sources easily and efficiently. Of course, there would need to be some regulation for this to happen... good luck getting a bunch of policy makers around the world to agree on such things. That's why I think this may only be possible in a self regulatory sense by global digital communities.


    FYI I'm also an open source developer and I spent much time with that when I'm not poking a little with alternative theory. I have an ordinary day job as a programmer. Right now I cannot work as they temporarily closed down all the servers that we used. Anyway the freedom to do do what I like is more important to me than making a ton of money or chase grants to do research that stay in line with those in power.

    I dabble with coding out of necessity , but I am self taught and probably not nearly as competent as you. ^^
    That's really nice to read more people doing things for other value besides just a monetary one.
    I think it's the key to solving what I call the energy freedom paradox.

Subscribe to our newsletter

It's sent once a month, you can unsubscribe at anytime!

View archive of previous newsletters

* indicates required

Your email address will be used to send you email newsletters only. See our Privacy Policy for more information.

Our Partners

Supporting researchers for over 20 years
Want to Advertise or Sponsor LENR Forum?
CLICK HERE to contact us.