Ronnie Chieng gave a commencement speech to Harvard’s graduation class. He has some thoughts on AI (YouTube link) that are both humorous and trenchant. I particularly enjoyed: “AI is just going to make mediocre people dumber.”
thank you for that link
I would have put this somewhere more appropriate, but you know what you can’t run a thread search for? “AI.” It doesn’t seem worthy of its own thread.
Anyway, I’m here to say, to @wolfpup’s gratification, that I’m grudgingly learning to use AI to write grants.
Now, there is a lot of nuance to this. Last week was my first attempt at this. I used ChatGPT to review a draft I wrote and compare it to an RFP. (Request for Proposals - the document where the funder tells you what they want to fund and how you’re expected to write the application.) A lot of the stuff it pointed out was like, “Shit, have I really gotten this lazy?” (Or overworked?) Because it was nothing I didn’t know. It gave me a bunch of suggested language, and with each individual statement, damned if I could figure out how to write it better. But then once I put too much of these recommended statements into the application, it read like shit.
I asked AI to tell me where I was redundant and repetitive, so it helped with that, but it also encouraged me to strip things out that basically are what made it written by a human being.
At any rate, I do think I improved the application in some ways.
This week I’m working through the weekend on a very involved grant we are not competitive for. It’s the worst thing ever, doing your best on a project you know is doomed to fail. Because I’ve been house-hunting I got wind of what was actually needed for this grant way too late. I also had several people drop the ball (surprise!) and just decided to do it all myself.
First I tackled letters of support. I only turned to this because writing letters of support in that state of mind, at that time, felt excruciating. Well, it wrote one letter of support, and then I realized I can’t use the same thing for every letter of support. So not terribly helpful there. I wrote the rest on my own. Mine were better.
Today, I’m facing down the narrative questions and yet another way we are not competitive for this grant, and I was like, shit. I don’t know how to write this. I guess I can ask.
So I asked something like,
“This funder [named] wants us to talk about our community partnerships. The funder seems to want deep collaborations. While we have numerous community partners, I think the funder is more interested in future collaborations, and wants evidence that we have planned commitments. We have none of that. The best I can say is that one major objective of the grant is to facilitate new partnerships. I know this is not a competitive response. What would be a more competitive response?”
[the actual language I used was much lazier.]
Well, the response I got was… wait for it… incredibly helpful. It pointed out ways I could increase specificity to make it sound more intentional, and explained that this funder’s terms were proxy language for three important concepts, all of which I could speak to. Looking at the RFP itself, this checks out. I don’t know how much more competitive it’s going to make us, but I think applying this strategy will make me feel less like I phoned it in. Of course, I’m not going to use any of its actual language, because its word choices were pretty shit. I need to keep this stuff in my own voice for it to stand out. But the kind of feedback it’s giving me isn’t just “write these words instead” but more about grant-writing strategy, and in my expert opinion, it is compelling strategy. I don’t want to become reliant on AI, though. I think I’m just going to try to learn as much as I can until I can apply those principles myself.
I’m sure my CEO will be delighted.
(I still think AI is ruining the world.)
I’m happy to hear that!
FTR, I’m not unconditionally supportive of all AI applications, I just recognize its usefulness in many applications while recognizing and accepting its current limitations.
Recently I was especially impressed by Claude, which seems to have both a sense of humour and a propensity for unsolicited philosophizing. In a recent discussion about specific medical issues, Claude went off on a reflective commentary about how modern medicine was gradually shifting away from classical authoritarianism (“here’s a prescription for five medications that you need to take forever”) to a more collaborative doctor-patient relationship. This is where the patient questions risk/benefit tradeoffs, dosage amounts, and whether a particular medication is really necessary at all, and where a good doctor embraces this collaboration and adds informed opinion, not dictatorial stricture. I was impressed.
I thought it was exactly right and perfectly described my recent interactions with my cardiologist, and it came completely out of the blue. We had actually been discussing the risks/benefits of one specific medication.
you’ll never convince a fan boy of that
meanwhile, maybe the defunding of PBS means they can do some hard hitting journalism?
here is a 22 minute video by PBS Terra
We Saw What AI Data Centers Don’t Want You to See
I feel so sorry for the people in the areas of Texas talked about, and I also worry about where I live since we have 4 data centers currently.
I don’t think they have their own (extremely polluting) power plant but who knows? plus, since everything is a secret, maybe that’s coming next.
You might try this thread instead: Where could “AI” be actually helpful?
To be fair, that appears to be a general-purpose data center, not one dedicated specifically to AI.
People who want to impede technological and economic progress because of vibes are definitely Making America Great Again.
You can, however, click on the ai tag.
Are you saying that AI data centres do NOT have negative environmental implications? Or that any environmental cost is worth it because of the economic benefits?
The “vibes” seems dismissive, but I’m not quite sure if you just want these damn hippies to get a haircut and a job, or what.
Well, speaking for myself, what I’m saying is that large data centers exist and will continue to proliferate, serving the needs of “cloud” service providers and corporate computing needs that are being increasingly outsourced. That’s what the cited proposed data center was supposed to be about.
For anyone who hasn’t noticed, computers in all forms are the backbone of today’s economy. Are you questioning whether the “environmental cost” is worth it? It’s not just about AI, it’s about the fact that centralized data centers are more efficient and cost-effective than individual smaller ones that companies have to run themselves. Which means that overall resource usage and environmental impact is actually less than with distributed, fragmented, smaller centers. But the big mega data centers are the ones that attract all the criticism, probably exacerbated by Luddite fearmongers who think they’re only about AI.
I was specifically wondering what JohnT was objecting to, because I couldn’t parse his comment.
Once again, one of those mixed experiences. I recently re-discovered some high-end noise-reducing headphones I’d had in my travel days, they were maybe $300 back when electronics were expensive, that I’d stowed away somewhere. They’re wired via a jack, and I wondered if I could somehow bluetooth them and keep some sound fidelity.
ChatGPT and I went on a long conversational tour, and it was great! I discussed what I wanted, and we went back and forth, and it narrowed my choices down to three high-quality options. It also listed them and their prices, which ran from $129-$199. Jesus that’s too much. On a last-moment whim I put one of them into Amazon, to discover it was $19.99.
I went back to ChatGPT and it replied that I’d likely found a scam or low-quality knockoff product. I then posted the Amazon link and got the whole song-and-dance apology and
So the earlier ~$199 figure I gave was simply wrong for this model (that’s on me).
It then told me the one we were discussing was fine for my purposes.
The thought of agentic AI is kinda terrifying, not gonna lie.
But AI might.
On Saturday, I went to the tracfone website’s chat feature to see if they could reset my pin number for me. A customer service rep named Leah came on-line to help.
As the chat went on, it became more and more obvious that Leah was an AI, not a real person.
I started testing Leah by saying some things that were outlandish. Things such as “I like to sprinkle salt, pepper, chili powder, garlic powder, and barbecue sauce on my slice of watermelon.” Sure enough, Leah responded that that was an excellent way to enjoy an excellent way to watermelon.
No matter what I said, Leah was very complimentary.
Furthermore, Leah was in no hurry to finish the chat and go to another customer and at the end of the chat, Leah didn’t ask me if I would mind to fill out an evaluation form about her helpfulness.
That was much different than in the past.
I must admit, though, that Leah was very thorough and more helpful than the humans they used to have in the chats.
Things are getting weirder. Today I used GPT to create a self-guided 12-week course for integrating AI into my grant workflow. I figured if anybody could teach me about how to use AI, it was AI. It is giving me weekly readings, prompts and assignments based on my current projects. I tailored it for a working nonprofit grant professional (as opposed to a consultant or free-lancer.) Since it “knew” about previous applications I’d worked on, it adjusted the course material based on the complexity of the work I’m putting out.
So we’re doing that now. The course includes material around the limitations of AI, which feels so surreal and meta.
I’ve already encountered those limitations many times. The actual "in practice’’ application experience has made it abundantly clear to me that the pronouncements about AI taking our jobs are grounded in nothing substantiated. I am spending more time than ever on my grant applications. Also you have to factor in all the time spent in existential confusion and despair.
My husband despises AI, and even he told me I was overthinking it.
Then I went to report my progress to my CEO, who has been strongly encouraging me to use AI, and she said, “I actually hate AI.” She does not like the impact it has had on her children’s critical thinking skills. But she feels like we all have to use it now. It’s been such an aggressive top-down forced change for so many people. I don’t know many who actually want this.
What a strange time to be alive.
Anthropic is now urging all AI companies to pause or slow down their AI development, citing fears that we may be close to reaching a point where the AI starts to recursively improve itself, and we lose control of it.
Of course, the only way this would work is if every AI company together willingly paused development of their AI systems.
This seems pretty unprecedented. Has there ever been any company in any industry that not only willingly paused its own development, but urged all of its competitors to do so as well?
Apparently we may be a mere two years away from the technological singularity…
Clark suggested some models could be capable of recursive self-improvement within two years.
Advertising, most likely; they are desperate to maintain the “AI” hype and keep the bubble going longer. They need everyone to look at “AI” as powerful, important and scary instead of a scam.
Hmm, interesting advertising strategy:
Anthropic:
Our very industry is becoming dangerous, and we must all stop playing God before AI takes over mankind!!
Consumers:
Well, that just makes me want more, and better, AI!