Home Creators Posts Import Register Favorites Logout

Videos

  • 575640261.mp4

Downloads

Content

ok i'm ALMOST feeling settled in this streamer habit---i didn't like the viewing experience of the last one where I just posted the stream recording as is, so this is my attempt at a happy medium between a livestream and an edited video. still gotta figure out actual streaming software so i can show chats before the next one but then i'll have this office hours thing in the bag~

also for anyone who watches it im particularly curious on yalls thoughts about the AI drama thing that starts at about 24:15

submit a question for the january office hours stream here

Files

Comments

Preethi Basani

Thanks for the very helpful answer! I actually did the LA program as well in my undergrad at Boston university. I did build connections with a lot of the scholars in the physics education group at BU so I’ll reach out to them! I also am still doing my phd( have not yet completed it) so maybe I can see if I can work with someone on a paper at some point before I graduate.

Preethi Basani

Also curious what the other ways to be socially engaged in a research context specifically that you mentioned briefly!

schlltchrn

Getting off Spotify is way less scary than it seems, I was postponing it too, but was surprised how easy it was, and it is worth it in so many ways! I think getting off streaming music is more of a shift, but not impossible. I know a lot of people moving from Spotify to Tidal, which is less horrible, and I recently moved to Qobuz, which is less horrible too. In Qobuz you can also transfer your playlists and liked tracks from Spotify directly. Otherwise, there are third party apps where you can do that for 5$ or so. Also, it surprised me how BAAAD the sound quality on Spotify is, after switching to better quality sound. It is beautiful. :) And all the music I listened to on Spotify was on Qobuz too, even though I heard that its library is smaller, sometimes you just have to manually search for it. Streaming is for sure not the most ethical way to listen to music anyhow, yes physical media is great, and also going to small local live gigs and buying merch directly from the musicians when possible is good too. However, I understand not everybody cares about music that much, and we are all living in a moral glitch, so not funding AI German military drones is where I draw my line right now, personally. I'd encourage everybody to look into it. Oh and thanks for a great office hour!

tatiana

did my question get skipped or was it not submitted properly / in time?

Craig

I'm listening to music much less these days, but it's been a huge part of my life. I've never had Spotify for the reasons you allude to regarding support for musicians and the damage Spotify has done to their livelihood. But I didn't know about the other stuff regarding military drones. It's almost like these billionaires know something we don't, with these "financial instruments" they're parking cash in.

Mark

Hi, Mark here! After hearing you read my question, I understand that my question implied that using an LLM is shameful. I did not intend this implication and I appreciate that you called it out, so I'll try to be more thoughtful and nuanced with future questions. While I personally do not like using LLMs, I agree that shaming individual people for using LLMs is not constructive and a distraction from more impactful action. I am sad to hear about what's happened since your discussion with Michael; we need to be able to have safe, good faith conversations! I hope things get better. For some quick background, I think you previously talked about the use of machine learning in astronomy for classifying stars or other celestial bodies. That sounded like a powerful and appropriately scoped use of large-scale-computing technology, so I wrote my question partly because I'm curious about reasonable uses of LLMs in an academic or workplace context. If you happen to think of anything (even if it's just a brainstorm), and you feel safe talking about it, I would appreciate hearing your thoughts. And if you're looking for other safe spaces to have these conversations, maybe Ali Alkhatib's "AI Skeptics Reading Group" would be a place to start (though caveat that I'm just aware that this group exists, I've never participated in it). Finally, on the topic of energy efficiency, I had also heard that DeepSeek uses less energy than ChatGPT. Then I came across this paper on arxiv, "How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inference", which includes a PowerBI dashboard comparing many popular LLMs on various sustainability metrics. DeepSeek R1 is at the top of the Average Energy Consumption (Wh) table at 64.31 Wh, whereas GPT-5 nano (minimal) is at 0.54. I'm not sure why this is the case, but after skimming the paper it seems like "hardware and data center efficiency, not model design alone, drives real-world energy use." If you have read this paper, what do you think about the methodology and the findings?

Dr. Fatima

didn't make the cutoff for this one, will answer in the next one

Xuanzi.Vyaiskaya (選子•エレVya)

Thanks for looking my name up!! :3 « Shwen Dz(i) » is probably the best rendition for English!! (Xuǎn zì = 選子, in Mandarin !! ) ( Xunzi tho — a famous philosopher from like 300 BC. XD maybe this was also me, my memory isn't too good xD ) Hmm. I'd tried to make it specific.. I guess i failed office hours 😭☠️.. 💀 Let me try to rephrase it, cuz like, I think what I was trying to ask was a specific thing. (Also, Alchemy/chemistry/particle physics weren't separate yet T.T ) Basically, if you had to do a promotional on a preEnglightenment person (or advancement if a person is less than ideal as a choice) whom would you pick? Hopefully this is better phrased 😭😭😭

Xuanzi.Vyaiskaya (選子•エレVya)

Thanks for posting these to look into!! I remember back when Playlist.com was a thing (Spotify uses Playlists form factor almost identically even.)

phil e the theyby

I haven’t finished the non-Patron questions part yet, but wanted to share thoughts on the “AI drama” thing (since there was a question about it). After seeing the show in SF, I was struck by how powerful your incorporation of your personal experience was for the presentation. It was the most major departure from your videos (where you usually don’t go into personal stuff in the same way), and I thought it was SUPER impactful. I also had the reflection afterward, if I were in your shoes, I 100% WOULDN’T put that sort of personal reflection and vulnerability into a YouTube video the same way I would a (non-recorded) live show. Putting your deep personal experiences onto the internet for the no-investment comments of random strangers is just SO likely to lead to awful shit. It shouldn’t, but it does, because the internet makes it vastly more difficult for folks to treat people on the other side of the screen like actual humans (especially when those people are women and/or BIPOC). When Michael posted his two videos, I was not super surprised by the negative reaction, because I had seen the same type of response to Alexander Avila’s videos on AI where he was actually being thoughtful and nuanced in talking about it. I felt kind of bad that your experience was dragged into it; I’m glad to hear you didn’t follow it (or know about it) as it was happening. The reaction largely wasn’t focused on you (from what I can tell), but it still felt gross. I think that, in terms of talking about AI, I’m less drawn towards saying “don’t talk about it in a nuanced way at all”; I see it as more, “anticipate the audience reaction.” There are people who are watching and listening who are thoughtful and are taking your points to heart; those people are unlikely to comment, whereas people who get out their pitchforks whenever they hear anything about AI that isn’t about how awful it is will ALWAYS comment. I think avoiding the topic where you may point to positive or understandable uses of AI in Office Hours is probably a good idea, since in that setting you don’t have the time or space to decide how important your response/comment is. So yeah, I’d probably avoid riffing on the topic, because there’s no good way to do a thoughtful cost-benefit analysis. Honestly, I feel like that kind of reflection on topics the internet struggles to engage with in good faith are probably best saved for live shows, where the audience has invested enough to be incentivized to truly reflect deeply on what you’ve said instead of just responding to hear themselves talk and “feel right.” Regarding making a video about the AI “drama” stuff, I don’t feel like that’s particularly aligned with how you’ve approached your work so far. You engage thoughtfully with comments and questions that seem to be offered in good faith, but I’ve never seen you as trying to read, anticipate, and respond to a kind of vague, unformed internet mob (and I’m not aiming to dismiss criticism people have, but comments that don’t engage with Michael’s appeal to humanity and empathy aren’t genuine critique, they’re knee-jerk online reactions). Your work brings out the thoughtful, better parts of people, and turns away from (very really opportunities to) tearing down people who have bad ideas or don’t engage with things in productive ways. To me, any sort of “response” video feels like it moves away from the growth-oriented, collaborative, thoughtful approach you’ve cultivated on your channel, and moves more into the amygdala-driven content fest that is so much of the rest of the internet. I don’t say that to downplay how fucking gross and stressful it is to feel afraid of speaking candidly because of how the Internet may react. But I think your boundaries (including in this Office Hours video!) have served you and by extension your work very well, and I think not having to care for or respond to the audience’s emotional outbursts is a good boundary to maintain. That said, I don’t fully know what you imagined that video about the “drama” looking like, so I may be completely missing something. The eternal caveat that “I don’t know you and do what works best for your soul” always applies :) Also that last Patreon question discussion was really awesome and rich, and I loved reliving that part of the live show’s discussion

Matus Chochlik

as the lead of a SW development team of ~10 people, (we all have recently been strongly pushed by the C-level management to start using LLMs for the explicitly stated purpose of increasing productivity) my approach is this: I have tried using LLMs and found them very inconsistent and unreliable for anything advanced, but then I have more than 25 years of experience, so I stopped using them completely. But then I can see that people with medium-level experience can get some advantages out of using them. So the use of LLMs is not banned just mildly discouraged and I mostly advocate for my position by going through all the wider impacts of LLM use. Yes it can in some situations help you do your work more quickly, but also: - it is unreliable and you have to check the output constantly, which usually is less enjoyable than making the work yourself. - environmental and resource usage impacts - training on stolen or exploited work of other people - gradual loss of your skills - getting more and more reliant on LLMs - etc. I've seen that focusing of the bigger picture made my team members start thinking more critically, than just talking about the technical merit or reliability of the LLM output.

AJ Biffl

inb4 chatgpt update: "aw shucks, sorry to hear about your eviction! here's a list of ten tips for building shelter out of garbage... no, the terms of your lease don't technically allow for this eviction, but remember landlords are people too, and everyone makes mistakes!"

Iris Purpurea

Also though, AI nuance aside, I’m in NYC and the vandalism of the friend AI ads on the subway has been so fun to watch

Wami Ogunbi

Great office hours video! And Phil e the theyby (what an iconic name, by the by) already wrote an excellent response regarding AI. I share many of the same opinions, so I will not repeat them here. I did want to highlight the point about how many people watch your nuanced videos, take the ideas to heart, and simply never comment. Meanwhile, a very vocal minority of people who seem to thrive on “drama” can easily fill up the comment section. I have watched all of your videos, and I initially commented on every single one as they came out. I have not been as diligent recently. I am now a tired, overworked college professor, and I just do not have the bandwidth to leave the thoughtful responses that I would otherwise love to write. I also find the debate about how to engage with “the other side” to be genuinely intriguing. At the risk of being misunderstood, I really think that our society has forgotten how to have productive debate. I see so much talking “at” each other from both sides. It is exhausting to witness even as a spectator. It feels like watching two soccer teams digging trenches on their own sides of the field, never touching the ball. Nothing gets accomplished, and everyone just ends up resenting each other even more. I do not think it should be controversial to want to have good faith discussions with people you fundamentally disagree with. Yes, I understand that a lot of hurt has been inflicted. But if the goal is to make our society better, to make the world better, then we cannot hope to achieve that goal if we refuse to listen to each other. If we stay focused on what makes one side “evil,” we will never be able to build anything together. And we do need each other if we want to create something better. The only alternative is one side dominating the other, and I do not see a moral or ethical way for that to happen.

phil e the theyby

I had forgotten about the comment on Michael Burns’s video about you being brainwashed by tech bros because you live in the Bay Area, that was so fucking funny

David Lightbringer

IMO, there are no ethical uses of LLM's because they - are all based on theft - are proven to increase psychosis, even with "normal" use - are designed to financialize what used to be human-to-human interactions - make you stupid by thinking / researching for you

David Lightbringer

That's not to say they cannot possibly be useful, but they cannot IMO be ethical given the context they exist in. There are however decent ways to talk to people about this stuff - the difference is "are you approaching someone with empathy or condemnation." I try to encourage people not to use them, for all the many reasons, but lots of people just want to use moral high ground to scold people and feel better about themselves, which is not based in empathy. That's just typical holier than thou stuff.

David Lightbringer

I do think the search for "ethical use" of AI tech is basically misguided. It's okay to identify the fact that the technology is brought to us by people with bad intent and that the effects of using it are so clearly harmful and just say "better not to use that." The #1 thing that AI tech does, functionally, is attempt to divert human-to-human interactions into the dead-end of an AI algo for the financial gain of the people who own the AIs. AI also greatly facilitates the top-down control & censorship of information, news, and culture, which I think should be a huge concern

David Lightbringer

I think it's tremendously positive for you to share your experience, particularly how it sucked you from one use to much broader use. Why would anti-AI advocates (which I am one) object to that? That's powerful testimony IMO

Raphaël Royer

Honestly, I appreciated your testimony so much, personally, I hate AI with a burning passion, I write reviews for new metal albums and the amount of album arts done by AI is shockingly high. So yeah, this small stupid reason is a big reason for my disliking, but probably the whole destroying of the planet and AI psychosis are maybe bigger reasons... All that to say, the fact that you had that expirience, it makes me fear it even more! and yeah when people tell me they use a chatbot, Im not an asshole about it, be empathetic, be normal!

Ray S

RE: AI and LLMs. Nuance is necessary but unfortunately no one wants it. My brother-in-law is obsessed with AI. He thinks it can and should be used to do anything. He thinks you can use a chat bot as a therapist, he creates AI slop as birthday and Christmas gifts for the family, and he kind of ignores all the social/economic/environmental issues with it. This is one end of the spectrum. On the other end, those who judge AI as 100% bad because they don't like the way it is being implemented are not helping. Data centers are exploiting towns, models are trained using copyrighted material and the creators aren't being compensated, people are using it in inappropriate ways. These are all real problems, but they are tractable. The answer is to understand the technology and close the legislative gaps that allow this exploitation and abuse to occur without unilaterally condemning the thing. (I think this is exacerbated by our geriontocracy, but that's a whole other issue.) Every technological advancement is messy and disruptive in its infancy. We need to get better at detecting, evaluating, and guiding these endeavors before the industry gets too far ahead of regulation. AI wasn't the first and it won't be the last.

Ray S

100%! Conversion isn't the play. In a pluralistic society we have to be able to allow others to be and think differently. And I think lots of people get that. The ones that are problematic are so because they don't realize we live in a pluralistic society. They think anyone who is different from them is an interloper in their world rather than the reality that we all share this world together.

Ray S

Sounds like the paper you are reading in journal club agrees with me based on your recent post. (I'm new to your community, so I haven't read it yet myself.)

Lipk

My youngest kid (13y) has been pushing me to buy a cd player. I actually resisted getting spotify etc for a long time, but at some point I caved and now I'm used to having my music accessible via phone. So I'm at the point I was afraid of years ago, locked in a service of this evil company, some artists I like are already leaving the platform. I think it's kinda funny that my kid is calling me out on it xD My oldest two don't get why I resist using eg google pay and amazon, but I think it's getting ever more clear it makes sense to resist getting dependent on these moloch's services. So I guess I'll go ahead and buy that cd player, and find my box of cds

Ray S

Might I suggest getting an MP3 player instead? Personally, I use a HiFi Walker H20 Pro with a 128 GB microSD card. This has 2 main benefits. First, physical media is bulky. Remember having to lug around binders full of discs to take your music with you? You don't want that. And you don't have to go 100% physical media in order to escape streaming services. Second, if you're using an MP3 player, you have some flexibility around where you get your music from. You can rip a CD, buy an album DRM-free from digital stores (e.g. Amazon Music, Bandcamp, Quboz), or pirate music (if you're into that sort of thing). In my opinion, MP3s are the best of both worlds: portability and independence.

Ray S

I think this is the crux of the whole argument about their utility. When I use a chatbot in a domain that I am an expert in, then I can clearly see when it is wrong. But when I am using it in a context where I am not an expert, then everything it says seems plausible and is often close enough to correct to be somewhat useful. So it is no substitute for expertise, but in the absence of expertise, it's better than nothing, like Dr. Fatima dealing with her landlord. I think the problem we face is that people might prefer to use a chatbot in private rather than ask an actual expert, thereby admitting to someone else that they aren't able to handle a situation on their own and also exposing themselves to the possibility that an expert might tell them something they don't want to hear but need to hear. The allure of AI is that you can maintain a facade of independence in the face of your shortcomings and you also get to keep your biases because AIs are sycophantic (at least today's AIs) and will avoid telling you you're wrong if at all possible. You get to live in this fantasy where you can pretend to be competent in everything and never wrong.

Atmosweeper

Thanks for the answer, even if you had to steer clear of addressing some parts of my question out of self-preservation! My knee-jerk reaction to your response about building student relationships was that you'd misinterpreted what I was getting at. Returning and reflecting a few weeks later, I think the bigger issue is that I wasn't ready to hear the answer you provided. To clarify, when I asked about student emails and rec letters, I was thinking more in terms of "please help me write a rough draft" or "please fix my tone" or "please make my convoluted wording more legible to a first-year undergrad." I sometimes get overwhelmed with student emails and lose my patience, but obviously, responses like "ffs the answer is in the LMS" and "please don't make me explain this again" serve nobody. For rec letters, it would be more like "Here's my unfiltered thoughts and talking points. Please convert them into a rec letter draft." I have no interest in having students interface with a chatbot instead of me. I suppose, however, that it could be a slippery slope from my intended use into a habit of glancing at student emails only cursorily, then accepting whatever the AI churns out with minimal human oversight (I think Technology Connections used the term "algorithmic complacency"). It took a while for it to sink in, but I really appreciate your point about the value that resides in the inconvenience of doing things the old-fashioned way. Thank you! I won't request further follow-up right now. I would love to hear more of your takes on AI once ideas have had a chance to marinate, but that's up to you to decide based on your judgment, comfort level, and interest. I absolutely support your steadfastness in maintaining your boundaries despite working in a profession that incentivizes creators to have none.