Show newer

"One of the fundamental questions in biology is how the vast diversity of morphologies in organisms is generated. Today, it is well-accepted that similar genes fulfil similar functions in different organisms. However, more than 40 years ago it was still assumed that morphological differences between animals were controlled by very different genes."

nature.com/articles/s41580-023

I'm now thinking that we will be running language models with a sizable portion of the capabilities of ChatGPT on our own (top of the range) mobile phones and laptops within a year or two

Show thread

Pathways to cultural adaptation: the coevolution of cumulative culture and social networks biorxiv.org/content/10.1101/20

'Now, scientists have discovered that Leonardo did detailed experiments that sought to illuminate the nature of gravity a century before Galileo and some two centuries ahead of Newton’s making its investigation an exact science. The scientists’ study of his gravitational ideas and experimentation was published earlier this month in the journal “Leonardo.”

nytimes.com/2023/02/17/science

Gotta admit that my first reaction to ChatGPT as a college instructor was a mixture of fascination & dread. Now I am encouraging my journalism students to use and interrogate AI services as part of their education. Here's why:

thescoop.org/archives/2023/02/

I accomplished a new milestone in my reverse engineering today. A colleague asked me to figure out how the “getfw” tool used in some Cisco images works so he could use Python to extract it.

So I threw it in IDA, narrowed in on a function called “fwdec” then dropped the assembly into ChatGPT … wait wut?

alperovitch.sais.jhu.edu/an-ex

My good friend @jags recently showed that ChatGPT is extremely useful for RE newbs like me so I ran with it.

It was able to explain the assembly code was loading strings into memory and those strings?

OpenSSL decryption commands; including the passphrase. Worked like a charm once my colleague plugged it into his code.

Yeah, it’s not reversing stuxnet but considering it took me - with nearly zero IDA skills - under an hour to figure it out I thought it was pretty damn cool.

@ncweaver @benjaminwittes Not only that -- but Google is already advertising their stress on "safety" and "reliability" as selling points for the as-yet unreleased Bard, which serves only to emphasize the inevitable flaws. Galactica, Facebook's ChatGPT-style thing, had ChatGPT-style problems (most notably, lying like mad rather than say "I dunno"). But because Facebook had marketed the thing as a portal to "scientific knowledge", that blew up so badly for them that they had to take it down.

So OpenAI just released a detector of AI-generated text, I assume because of concerns in education / homework.

openai.com/blog/new-ai-classif

Maybe this is good?

No, it's very bad.

They claim 26% true positives, 9% false positives. Assume 10% of submitted homework is chatgpt generated, you get the classic counterintuitive outcome of poor predictive power: if a homework is flagged, there's a 3:1 chance it's *human* generated.

This is going to cause a lot of harm. It should be immediately recalled.

And in case this post wasn't clear: I'm all-in on large language models: they confidently pass my personal test for if a piece of technology is worth learning:

"Does this let me build things that I could not have built without it?"

What I find interesting is that - on the surface - they look like they solve a lot more problems than they actually do, partly thanks to the confidence with which they present themselves

Figuring out what they're genuinely good for is a very interesting challenge

Show thread

Humans adaptively select different computational strategies in different learning environments. biorxiv.org/content/10.1101/20

The physics big science wish list

"The future belongs to those who prepare for it, as scientists who petition federal agencies like NASA and the Department of Energy for research funds know all too well. The price of big-ticket instruments like a space telescope or particle accelerator can be as high as $10 billion.

And so this past June, the physics community began to consider what they want to do next, and why."

nytimes.com/2023/01/24/science

2023 stem cell & regenerative medicine predictions ipscell.com/2022/12/2023-stem- FDA, biotechs, cloning, exosomes, CAR-T, embryos, CRISPR, more

It’s very funny to me that the dominant Twentieth Century conception of AI was a slightly awkward nerd with an inhuman mastery of facts and logic, when what we actually got is smooth-talking bullshit artists who can’t do eighth-grade math.

Show older
Qoto Mastodon

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