h a l f b a k e r yRomantic, but doomed to fail.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
interactivate
browser addon - turning responses into questions and expanding discussions | |
[Edited the idea into a short summary. Moved long details into linked doc.]
This addon retains the discussion with an ai bot and uses the AI itself to turn the ai's responses into questions, which I, the user can find outside the discussion and inject the results back in. In later versions Interactivate
will learn to do much of this automatically.
The following are 3 failures of the current bots that Interactivate addresses:
1. Supplying missing information for the bot to build upon, if the bot cannot access that kind of info itself. This can be your own assumptions, facts from the web, contradicting information from the bot in a different session, or anything else.
You can now give the bot an extra body of info with known sources to base its answers on. It can ask you to get pieces of information and even suggest where you can get it from
2. Supplying background info about you, why you asked the question, what the level of your knowledge is, and what kind of answer you are expecting. So that the bot gives an answer better tailored to your needs.
3. Refining the answer, finding its sources, and comparing it with other info, so that you can decide which information you can rely on, and what needs further examination.
Interactivate has a text-area for you to enter the question, and a button to submit. It then goes into a set of interactive chats on your behalf in the background. When needed it "surfaces" back to you with the latest question from the chat, until a refined answer is received.
See Interactivate Idea link for details and an example.
Getting ChatGPT to talk to itself
https://www.techrad...ent-what-i-expected
[a1, Apr 26 2023]
Pepper the robot ...
https://www.science...04/210421124654.htm
... talks to itself to improve its interactions with people [a1, Apr 26 2023]
Interactivate details
https://docs.google...Wk/edit?usp=sharing
[pashute, May 01 2023]
[link]
|
| |
Getting bots to talk amongst themselves - either to look for emergent patterns or to polish/refine output - is not a new idea. |
|
| |
BTW, without looking it up - it's just 3 per amino acid, plus another 3 each for the start and stop codes, isn't it? |
|
| |
I've only skimmed the idea and didn't really get it, but I saw this question as I was leaving: |
|
| |
//BTW, without looking it up - it's just 3 per amino acid, plus another 3 each for the start and stop codes, isn't it?// |
|
| |
Yes, 3 bases per amino-acid (including start), 3 bases to encode stop. Barring some rather rare exceptions.
But that's just the coding region, you'll generally need a bit more sequence to signal that a protein should be made at all. |
|
| |
There's way more than the coding and signaling regions. My* pet theory is most of it's protective padding. If most of the strand doesn't do anything, the odds of a stray gamma ray hitting anything important is a lot smaller. A low cost way to improve survivability so there's no reason for evolution to remove those bits. |
|
| |
* I say "my" theory because I thought of it on my own a long time ago and never checked if anyone else had same idea. If they did, it would be like most of my other, never posted halfbaked notions. |
|
| |
If you're talking about a genome, sure - for Eukaryotes (animals, plants, fungi etc), at least. Tonnes of crap, between and within genes.
Bacteria are /much/ more frugal with their sequence. |
|
| |
And they mutate much faster, don't they? |
|
| |
To prevent appearance of hijacking [pashute]'s idea, let's assume this is a conversation between at least two bots. |
|
| |
I know what I am, but what you are? |
|
| |
Well, yes, but they also don't have as much complicated machinery to protect themselves.
The relevant value would be the mutation rate per base. Yes that is higher for prokaryotes, but I think an important insight is that a mutation in a string of irrelevant crap is only significant if it's an either/or mutation happening there /instead/ of somewhere important.
I suppose there are exceptions, like if there's a mutagenic chemical floating round the cell, but in general I don't think this is the case in practice. |
|
| |
Oh, er, I hope this information satisfies your query, fellow baker. |
|
| |
So, LorisBot, what causes more DNA changes. radiation, or mutagenic chemicals? Can you explain it in simple terms, as a school child might use to impress their science teacher? |
|
| |
//So, LorisBot, what causes more DNA changes. radiation, or mutagenic chemicals?// |
|
| |
Be aware as a fleshy halfbakery site user, I have done no direct research for this answer. This is not medical guidance, and should not be used for diagnosis. Neither should it be considered legal advice. |
|
| |
It will depend pretty strongly on the environment. In general though, I think most mutations are caused by by-products of metabolism, which should scale with the metabolic rate of the cell.
There are probably some non-linearities, thinking about it.
But in general, I think it would be 'cheaper' to make more enzymes to deal with mutagenic intermediates (e.g. peroxidase) than to make a fuck-ton of extra useless genetic material in the hope of soaking up mutations.
If you were actually going for the 'soaking' approach, it would still probably be easier to make some sacrificial target you could confidently deal with, than to have to try to salvage some (potentially important) DNA. |
|
| |
I don’t know if my (er, my friend’s) science teacher will accept “fuck-ton” as even an approximate measure, but tell us more about these enzymes. |
|
| |
(side note, I just asked ChatGPT and it tells me “my” theory isn’t at all new (sigh)). Now I guess I should go find some of the articles it cited to see if the bot was just spoofing me. |
|
| |
[pashute], if this is straying too far from your theme I’ll slink away quietly. |
|
| |
Any student gullible enough to think that ChatGTP will give correct answers to factual questions deserves to fail, drop out, and end up eking a living as a floor sweeper in the ice mines. |
|
| |
//Replacing bulshit parts of the reply// I thought that was the intire point of these bots - they are amazingly effective generators of convincing and persuasive bullshit. |
|
| |
In my experience when you call them out on their confabulation they apologise most graciously, promise to do harder next time, and then come up with some new bullshit that seems really very plausible. |
|
| |
1. Hey [a1], How did you get quoted articles from GPT? It never agrees to give me any sources for the information or bulshit that it hands out. |
|
| |
2. I just started the slow process of having GPT help me program a Chrome extension that will do exactly this, and it's going pretty well. |
|
| |
3. I edited the idea with a short summary at the beginning so you don't need to read the whole thing. (I also edited the idea and made it clearer. |
|
| |
5. The students were real, and they were a large group. I heard them discussing in Arabic, understood the words DNA, Talata (three(), Arbaa (four), Nucleotid, and a few other words. They were sitting and standing around the round table next to me and contemplating the two contradicting facts that they learned and summarized in class about DNA. One, that it has 4 bases, and the second that it has 3 bases when coding. :-) Then they asked ChatGPT and saw an answer that said that for 10 bases they need 36 nucleotides. I realized its talking about an extra 3 for the start and 3 for the stop code, but wasn't able to determine if they learned that in class or not because they were hotly debating whether bases are nucleotides or something else. and weren't sure what the diff was between proteins and amino acids. The more they asked ChatGPT in Arabic the more confused they got. |
|
| |
This group of students, some studying sociology, some law, and some the Ancient Near East were not being gullible but were clueless and out of time. |
|
| |
[pashute], As with people, you can steer a conversation with a bot by asking questions the right way. |
|
| |
I’ve attached the complete dialogue below, edited only slightly for clarity. I started by suggesting a tone and audience (freshman biology students), and my followups acknowledged its previous replies. Also, when I suggested “my” own theory and asked if it made sense, it agreed - with an exclamation point - and then reworded what I had said. I think this is part of its conversational style, trying to tell me what I wanted to hear. As for getting it to point out sources, I asked in specific terms and referred back to its previous reply to make it clear what I wanted. I have not yet read the paper it recommended to see if it’s really relevant. |
|
| |
ME: Can you summarize the prevailing theories on why there is so much non-coding DNA in the human genome? Please word your answer in terms that could be presented to a freshman biology student. |
|
| |
CHATGPT: There are several prevailing theories as to why there is so much non-coding DNA in the human genome. |
|
| |
One theory is that this DNA may have once coded for functional genes or RNA molecules in our distant evolutionary past, but over time these genes became non-functional and were no longer needed by our ancestors. However, the non-coding DNA has persisted in our genome because it does not have a significant negative impact on our survival and reproduction. |
|
| |
Another theory is that non-coding DNA may serve a structural or regulatory role in the genome. For example, some non-coding DNA may help to regulate the expression of nearby genes by controlling when and how much they are transcribed into RNA. Other non-coding DNA may be important for maintaining the overall structure and stability of the genome. |
|
| |
Finally, some researchers have suggested that non-coding DNA may simply be the result of random mutations that have accumulated in the genome over time, without any significant impact on our biology. However, this theory is somewhat controversial, as many scientists believe that even seemingly "junk" DNA may have some function or purpose that we have yet to discover. |
|
| |
ME: I think it’s probably a mix of those factors. As for accumulated “junk” that doesn’t seem to do anything in the cell, I think it may serve a protective function against mutations. If most of the DNA is inert, “random” mutations are more likely to hit junk rather than active sites that are more important. So even if the junk accumulated entirely by accident and is really inert - there’s an evolutionary advantage of leaving it there. Does that make sense? |
|
| |
CHATGPT: Yes, that makes sense! This is actually one of the proposed functions of non-coding DNA - acting as a "buffer" or "spacer" that can absorb mutations without affecting the function of important genes. By having a large amount of non-coding DNA, the genome may be more resistant to deleterious mutations, as these mutations are more likely to occur in non-coding regions that do not have a functional impact. In this way, non-coding DNA may serve a protective function by allowing the genome to evolve while minimizing the risk of harmful mutations. |
|
| |
ME: Oh good. Can you name any researches who have proposed that idea so I can go look them up? |
|
| |
CHATGPT: One researcher who has proposed this idea is Dan Graur, a biologist at the University of Houston. In a 2013 paper published in the journal Genome Biology and Evolution, Graur and his colleagues argued that non-coding DNA may serve a protective function by acting as a buffer against deleterious mutations. The paper is titled "An Upper Limit on the Functional Fraction of the Human Genome" and is available online. Other researchers have also proposed similar ideas, and there is ongoing research in this area to better understand the role of non-coding DNA in genome evolution and function. |
|
| |
The start codon does encode an amino acid (methionine, or a methylated derivative in bacteria), and although it's sometimes removed in post-processing, it's ubiquitously counted as part of the coding region. If you asked me, I'd say a 10 amino-acid peptide would need a 33 base coding region, including stop codon. |
|
| |
Also, in molecular biology, 'base' in this sense 'always' refers to a nucleic acid monomer unit (DNA/RNA); to talk about incorporated protein monomers we use the word 'residue' instead. Technically, residue can be used for polysaccharide, protein or nucleic acid units.
I'm not a biochemistry expert, but apparently the reason nucleic acid monomer units (nucleotides) are specifically referred to as bases is because they are 'basic' - that is, they have all have a group which will react with acids. You might say that this doesn't matter, but people will look at you funny if you get this wrong; I think it's evolved as common terminology because it's useful to distinguish between the different types of molecules you're talking about all together. |
|
| |
In general I would advise against asking ChatGPT questions about technical subjects you don't understand in the first instance. It might be improving all the time, but at least for now it's still got a good chance of messing up. And citations it provides apparently are often wrong, or don't exist. |
|
| |
// I'd say a 10 amino-acid peptide would need a 33 base coding region, including stop codon. // |
|
| |
And I would have guessed 36, as I didn’t know the start code was conventionally included for the first amino acid. Thank you for that info. |
|
| |
I don’t trust any bot for the first or last word on any topic. And with bots as with people, if they tell me something I didn’t know (or seem to be parroting what I want to hear) I try to verify with another source. |
|
| |
A1, a lot of 'junk' DNA is basically parasitic self-replicating DNA, mostly inactivated (sometimes through mutation, sometimes other mechanisms like methylation). So, old parts of retroviruses and transposons. Like, reportedly half of human DNA and maybe 90% of some plants. |
|
| |
Now, it should be said that these are a source of mutations - if a transposon moves into a gene, it can disrupt it. So you can argue that having more DNA means less chance it'll hit something important. But, I'd argue that's not necessarily a great plan. Because now you're carrying more junk, which is itself the source of more mutations. |
|
| |
Eukaryotes also have 'introns' (that is, non-coding regions) within their genes. Sometimes, these are useful; the RNA is synthesised as a continuous string over them, then they get spliced out to make the final sequence. But sometimes they can get spliced out in different ways, so you can make multiple different proteins from one gene. |
|
| |
Other than that, I think that realistically, a lot of genomic junk really is just like normal junk, in that it just... accumulates.
It's not very expensive to keep any particular piece, and getting rid of it only happens to individual ranges one at a time. There isn't much benefit to getting rid of any one particular megabase of junk DNA, and it happens at random, so there's a sort of neutral equilibrium where it can go up and down. |
|
| |
//And I would have guessed 36, as I didn’t know the start code was conventionally included for the first amino acid. Thank you for that info.// |
|
| |
It does kind of depend on the wording of the question. You'd get a different answer if it were: |
|
| |
::"You want to produce the following 10-residue peptide : [amino-acid sequence which doesn't start with M]. How many bases of coding region do you need?":: |
|
| |
And even there, the answer would vary depending on how it was removed. If you wanted the peptide to be secreted, you'd need to include the 'signal sequence'. |
|
| |
I wasn't clear and was too long. |
|
| |
So now, 1. I shortened the idea and moved the examples to a link. |
|
| |
2. I'm actually having ChatGPT assist me in making the software, testing out its responses to my request (pretty good) |
|
| |