Robin wrote:
> These bots have an API that lets them access wiki, or so they claim.
> Try asking for something else that is not on LENR-CANR.
I excluded all data outside of LENR-CANR.org. That is an option in the
FastChat Bot configuration. This Bot used the word "J'Accuse" as a synonym
for
In reply to Jed Rothwell's message of Wed, 17 May 2023 11:32:10 -0400:
Hi,
[snip]
These bots have an API that lets them access wiki, or so they claim.
Try asking for something else that is not on LENR-CANR.
>[The word "J'Acuse" appears nowhere in the LENR-CANR corpus.]
>
>. . . He also wrote
This is uncanny.
CHATBOT 2:
BOT: Hello! What can I do for you today?
What did Mallove say about the experiment at MIT?
BOT: Eugene Mallove accused MIT of publishing fudged experimental findings
to support their early condemnation of the work of Pons and Fleischmann. He
believed that MIT's
H L V wrote:
> For example, although chatgpt 4 is more accurate than 3, it has lost some
> speed.
>
That is typical of computer programs. You often have to choose between
speed and accuracy. When there is a tremendous amount of data, I often run
a program to extract summary data, to give a
On the other hand we have not started yet to optimize the speed of them.
Also consider what is happening on the hardware side.
On Wed, May 17, 2023 at 3:18 PM H L V wrote:
> A cognitive scientist has suggested that the performance of these chatbots
> may have operational limits or trade-offs
A cognitive scientist has suggested that the performance of these chatbots
may have operational limits or trade-offs that we do not yet understand.
For example, although chatgpt 4 is more accurate than 3, it has lost some
speed.
Harry
On Tue, May 16, 2023 at 9:36 PM Jed Rothwell wrote:
> The
That is just amazing.
On Tue, May 16, 2023, 9:36 PM Jed Rothwell wrote:
> The LENR-CANR.org ChatBot has been upgraded to ChatGPT version 4. It seems
> to work better, with fewer hallucinations. It may be slower.
>
> I asked it some of the questions version 3 could not answer. It answered
> them
The LENR-CANR.org ChatBot has been upgraded to ChatGPT version 4. It seems
to work better, with fewer hallucinations. It may be slower.
I asked it some of the questions version 3 could not answer. It answered
them correctly. It still gets confused. It thought that McKubre made a
statement that
Beware of hallucinations! This is completely wrong:
What are the author(s), title and URL of the Pd-B experiments at China Lake?
BOT: The author of the paper on Pd-B experiments at China Lake is Miles, M.
and B.F. Bush. The title of the paper is "Palladium-Boron Cathodes Produce
Excess Power in
I replaced all files with text formatted files, and corrected some that
were not machine readable. I think the ChatBot performance has improved. It
finds URLs more often, and it finds details in the papers. It still misses
some details. See this exchange:
BOT: Hello! What can I do for you today?
10 matches
Mail list logo
