Show newer

@TheStrugglingScientists I would add to this the ability to use hashtags for organizing and searching your ideas. Moreover, there is a tremendous amount of community plugins that let you adapt the system for different purposes (e.g. calendar, templates, daily notes, and functionality to automate certain aspect).

@golub_fdm @chemistry @nfdi @researchdatamanagement OpenBIS is used a lot within ETH Zurich as it was created here and support is excellent. It is free to use and mainly geared towards life science research. It was originally developed as a LIMS for omics research and the ELN was put on top of it at some point. It does not support chemical structures or stoichiometry, so it is not used by synthetic groups. It offers a lot of functionally for adding meta data and linking data. Numerous groups from pharmaceutical science, all branches of life science, biologically oriented chemistry and even from chemical engineering are using it.
openbis.ch/

ChatGTP, please write my a rhyming chemistry poem on failing reactions and frustration in the synthetic organic laboratory with a silver lining of a small success.

"In the synthetic organic laboratory,
Frustration often fills the air,
As failed reactions pile up high,
And our spirits start to despair.

We mix and we stir,
But the results are not what we desired,
Our hopes are dashed,
And our confidence is tired.

But then, at last,
A small success we do find,
And though it may be fleeting,
It helps to clear our mind.

For though the road to victory
May be long and hard to tread,
With perseverance and determination,
We will succeed in the end.

So let us not give up,
But continue on our quest,
For even in the darkest hour,
A glimmer of hope still exists."

Not too bad for something that took 3 seconds or so to produce.

Next generation of GPT bot rumored to so realistic that when you ask it to write something important to meet a deadline it goes ahead and cleans your house.

@ct_bergstrom @ct_bergstrom What I find stunning is the depth of the deception.... that the implicit effort that's put in to create a believable facsimile of a source is every bit as deep as the effort to generate an answer in the first place. It's babble, but it's exquisite in its detail... a Potemkin village where all the tiniest details are thought of.

@cgseife

That's a wild story. What a trip.

#Galactica did the same thing, and I found that profoundly irresponsible given its claimed abilities and purpose.

Even when the citations refer to real papers, the claims associated with them are often nothing like what the cited paper asserts. That's the problem with an LLM in this context. When it adds a citation, it has no representation of what claims are or are not in the paper it is citing. The citation is just more just convincing babble.

@ct_bergstrom It turns out that those citations are ENTIRELY MADE UP.

Even though one would expect that the ACLU and Wired would've written about the NSA's collection practices, as far as I can tell, those articles never existed; nor did the Times or Smithsonian articles on Clever Hans even though it would've been natural to see such a piece in their pages.

Even the URLs are constructed to look plausible. But they're total fictions. I think.

Very creepy construction of fake knowledge.

Playing around with #OpenAI's #GPT3 text generator led me to perhaps the creepiest behavior yet. As @ct_bergstrom noted, the AI bots are able to cite the sources of their own writing.

I decided to see if it could function as a plagiarism detector. But when I put a few snippets of my own writing in, it found several uncited sources. (Attached.)

Aha! Someone must have plagiarized me! Or so I thought. But the truth was much stranger....

Via @grwster an interesting observation to which I am still looking for a counter-example.

If you take text generated by #ChatGPT and ask it "Was this text generated by you or by a human", it correctly identifies all text it wrote.

It does, however, mistake text written by other AIs (e.g. lex.page) for human-generated text.

Anyone have a positive or negative counterexample of ChatGPT's ability to identify its own writing?

@ct_bergstrom Maybe we need to start setting exams where we tell students to use the cheat-bots and paste in the answers, but then the actual exam is to GRADE THE BOT’S ANSWERS and explain which answers are wrong and why 🤣

@sanjaysrivastava Congrats and all the best for what's to come. I started following you because of black goat and allways enjoyed it. Thanks for all that BTW.

Me: It's really sad to see how easily people are influenced by social media

Also Me: sees one tiktok on cheese cake muffins and goes to the store to buy ingredients

So, dear #chemistry #chemiverse, I have created a group called @chemistry. If you tag the account in your toot it should auto boost the toot, to see those boosted toots in your timeline, follow the group. Hope that makes sense.

Show thread

@susannegruss We were speculating this morning if this was actually invented way back by some parent who had trouble rousing their kids in the dark time of the year...

@susannegruss My kids got up more than half an hour than usual and at a bootup time that was less than a quarter of the usual!😆

In case you missed it, here's last week's comic illustrating the history of #Twitter in four panels:

Show older
Qoto Mastodon

QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.