lmao, Zoom is cooked. Their CEO has no idea how LLMs work or why they aren’t fit for purpose, but he’s 100% certain someone else will somehow solve this problem:
So is the AI model hallucination problem down there in the stack, or are you investing in making sure that the rate of hallucinations goes down?
I think solving the AI hallucination problem — I think that’ll be fixed.
But I guess my question is by who? Is it by you, or is it somewhere down the stack?
It’s someone down the stack.
Okay.
I think either from the chip level or from the LLM itself.
I think solving the AI hallucination problem — I think that’ll be fixed.
Wasn’t this an unsolvable problem?
it’s unsolvable because it’s literally how LLMs work lol.
though to be fair i would indeed love for them to solve the LLMs-outputting-text problem.
Yeah. We need another program to control the LLM tbh.
Sed Quis custodiet ipsos custodes = But who will control the controllers?
Which in a beautiful twist of irony is thought to be an interpolation in the texts of Juvenal (in manuscript speak, an insert added by later scribes)
Lol I like how they put the author’s note at the beginning of the article, “this was a very special interview” as if it’s special because of the unique insights instead of special because it sounds coked up.
Haha at the chip level? What’s he smoking?
What’s he smoking?
Whatever he’s smoking, it’s strength rating is at least: “make it seem like a good idea to call employees back from remote work despite that being the one thing we sell”.
So that’s gotta be some strong stuff.
the important thing about Zoom is that it was the lucky winner of the pandemic. Could have been Google Meet, could have been any of their other competitors, but somehow everyone just converged on Zoom.
Having worked in an IT department in 2020, it wasn’t just random. Zoom was stable for large meetings and scaled pretty smoothly up to a thousand participants. And it’s a standalone product and it had better moderator tools.
MS Teams often got problems over around 50 to 80 participants. Google Meet worked better but its max was way lower than Zoom (250?). I tried a couple of other competitors, but none that matched up (including Jitsi, unfortunately).
So if you were at an IT department in an organization that needed to have large meetings and were looking for a quick solution that also worked for your large meetings , Zoom was in 2020 the best choice. And big organisations choices means everyone has to learn that software, so soon enough everyone knows how to use Zoom.
They were at the right place, had the better product, gained a dominant position. And now they are tossing all that away. C’est la late stage capitalism!
MS Teams often got problems over around 50 to 80 participants
As honourable mention, MS Teams is also uncontrollable, overblown jank that
- doesn’t work in a browser despite being built in Electron
- is complete shite on Android, despite bring built in Electron
- barely works on Windows, thanks to being built in Electron but despite the fact that it’s built by the Windows people
And even at its best behaviour it randomly loses messages while eating up way more CPU and RAM than possibly justifiable for a glorified IRC UI.
No wonder Zoom won out over that one, if you tried to use Teams in 2020 you barely could.
Also according to my freelance interpreter parents:
Compared to other major tools, was also one of the few not too janky solutions for setting up simultaneous interpreting with a separate audio track for the interpreters output.
Other tools would require big kludges (separate meeting rooms, etc…), unlikely in to be working for all participants across organizations, or require clunky consecutive translation.
So… what are the chances Zoom ends up recording people in Zoom meetings to use as training data?
I am suspecting this is the real reason everyone is coming out with their own LLM. Pumping massive amounts of arbitrary data sources into them feels justified to the average user and it’s a convenient excuse to collect whatever they want without needing to invent “features” which justify it. Yeah thanks for showing me a picture I took 10 years ago snapchat if I wanted to see it I’d look at my photo album but I guess now you have a “reason” to keep it
pushing 100%, they already tried to sneak in a TOS provision to do just that
I believe that LLM can represent me anytime.
Talk about having the most replaceable job.
he admits it dot webp
AI is, roughly speaking, complete bullshit that should die in a fire.
But…
I work remotely and I would love if I could make Zoom make me look like my hair is perfectly combed, I’m wearing a collared shirt, I’m smiling, and in general I’m presentable and professional-looking without anyone on the Zoom call ever knowing it was a trick.
Not to say I think we’ll achieve that any time in the foreseeable future, but one can wish.
you can readily achieve that with video synthesis and overlay techniques from circa 2016~2017, tbh
(not saying you should, just saying it is entirely possible and has been for years)
Yeah, I’ve looked into it a little, and I know there are technologies for achieving that, but then again we’ve all seen videos of filters failing, yeah?
It’d either have to be so reliable I could depend on it not getting confused and messing up even once even over many years or have some failsafe that made it just look like there was lag or something rather than showing my real face. (But also it couldn’t glitch if I stepped out of frame. Like, it’d suck if it decided the lamp behind me was my face or whatever. And probably it couldn’t restrict what I could do. Like, I wouldn’t want to have to be careful about the angle at which I turned my head or anything.)
And it would have to be sufficiently plug-and-play that it wouldn’t take me a year of tweaking, coding, and testing to get right before deploying it for real.
So, I dunno. I haven’t experimented with anything like overlay techniques directly, but it still seems implausible to me that anything fulfills all of those requirements.
I suppose I’d also settle for “it glitches sometimes, but not in ways that show how uncombed my hair is and how rumpled my shirt is, but also everyone does it and everyone knows that everyone does it, so it doesn’t reflect poorly on you if they discover your secret.” But I think we’re also a long way off from that.
And to be fair, I’m probably overthinking it to an extent and it probably wouldn’t actually reflect poorly on me if I did it. (They might be impressed. Who knows.) But still.
Start using AI from the top…REPLACE the CEO of zoom with an AI, if it works…then it can be implemented downwards.
it’s pretty clear this has already happened. The Singularity Is Here And It’s Dumb As Shit.
what if the push to replace workers with chatbots was initiated by chatbot CEOs
I think this is a marvelous feature. Instead of attending useless calls or worse, in person conferences, yourself you have an unlimited amount of „digital twins“ that can accomplish nothing on your behalf.
FYI Jitsi Meet is an open source video conferencing app. It works similarly to Zoom, and is free. One can even host an instance on one’s own system.
“About Jitsi: Video Conferencing Software. Jitsi is a set of open-source projects that allows you to easily build and deploy secure video conferencing solutions. At the heart of Jitsi are Jitsi Videobridge and Jitsi Meet, which let you have conferences on the internet, while other projects in the community enable other features such as audio, dial-in, recording, and simulcasting.”
Personal Note: In 2020 I took my classed online, having chosen to use Zoom, tho’ Jitsi was on the list. Last September when Zoom changed their terms of service to include using our videos to train AI, I decided to switch to Jitsi. My operation is pretty small, and the $160/year I spent on Zoom was an expense I no longer needed.
I’m a fan of jitsi. That said, like a lot of open source I think it lags in features and bug fixes. Joining from Linux I have a lot of issues with my webcam not being recognized or my audio being very low for other participants, while I don’t have those issues with the remedial Linux version of their app that Zoom put out.
I still use jitsi, I just don’t feel I can rely on it for mission critical applications, unlike say, VLC.
What’s the elevator pitch now?
With conferencing, again, this is one app. If you look at your calendar, it is not only to join your video meeting but also a lot of other things. You read emails, send a chat message, make a phone call, have a whiteboard session, schedule something with external third parties. What we are doing now, it’s really looking at your entire schedule, how to leverage Zoom Workplace to help you out. Essentially, you can leave Zoom Workplace, and Zoom Workplace can help you get most of your work done, right? That’s our pitch.
That might be one of the worst pitches I’ve ever heard. Does that actually mean anything to anyone?
Unrelatedly, I need to quickly sneer at another quote:
like in 1995 when the internet was born
1995? How was the Internet born in 1995? Eternal September already started two years ago at that point.
Even Mosaic was 2 years old already in 1995, so the web—much younger than the infrastructure—was a solidly established thing. Did he just make up the number or is that some benchmark, I wonder?
Maybe the “corporate” web?
Q: When you think about the big vision — which still my mind is blown that this is your big vision, — of “I’m going to send a digital twin into a meeting, and it’s going to make decisions on my behalf that everyone trusts, that everyone agrees on, and everyone acts upon,” the privacy risk there is even higher. The security surface there becomes even more ripe for attack. If you can hack into my Zoom and get my digital twin to go do stuff on my behalf, woah, that’s a big problem. How do you think about managing that over time as you build toward that vision?
A: That’s a good question. So, I think again, back to privacy and security, I think of two things. First of all, it’s how to make sure somebody else will not hack into your meeting. This is Eric; it’s not somebody else. Another thing: during the call, make sure your conversation is very secure. Literally just last week, we announced the industry’s first post-quantum encryption. That’s the first one, and at the same time, look at deepfake technology — we’re also working on that as well to make sure that deepfakes will not create problems down the road. It is not like today’s two-factor authentication. It’s more than that, right? And because deepfake technology is real, now with AI, this is something we’re also working on — how to improve that experience as well.
Spoken like a true person who has not given one iota of thought to this issue and doesn’t know what most of the words he’s saying mean
“What if you get hacked?”
“Simply don’t get hacked.”
Spoken like a chatbot you mean. (raises suspicious eyebrow at Definitely-Human Notabot, CEO)
“the industry’s first post-quantum encryption.” What the hell is post-quantum encryption?
According to NIST this is something to be developed, not something Zoom has ‘all of a sudden created’ in the time between that question being asked, and the time the question was answered. SMH.
If you are curious, you can read up on it: https://csrc.nist.gov/projects/post-quantum-cryptography
(I realize other comments downthread have already addressed some of this, no slight to others intended)
so, PQC is definitely not snakeoil, and it’s actually seen uptake in a lot of things over recent years (just off the top of my head: openssh 9.0 in 2022, evolving work in implementations in TLS ciphers, etc (and as much as I fucking dislike cloudflare, they are actively funding a lot of forward-looking cryptographic work - thus being one to link to)). but as with all things cryptography, it’s a moving and changing field
the industry’s first post-quantum encryption
I suspect in this statement, “the industry” is load-bearing and inspecific, and resolves as “the industry of things that do what zoom do”. it is a highly vague statement though, and I 🤨 at it being used as it was where it was
(e: I did look up their actual announcement about this; “UCaaS” kill me)
I’m reticent to make any further specific claims/statements re the rest of PQC, since while it is one of my areas of interest and in which I keep relatively informed, I’m also not a cryptographer by trade and consider my knowledge at best armchair-competent. pretty damn interesting field though, if you have any interest in math or cryptography it’s well worth diving into it sometime :)
I thought we already had post quantum encryption, or at least that’s what some articles I read claimed
Please elaborate. I’m def not up on the cutting edge of encryption. And I’d like to know more.
it means cryptography with algorithms that will be resistant to quantum computers that are any good
Thank you, I understand the goal in a broader sense, and definition. Are you aware of any methods, for instance, that Zoom, or anyone else, could actually be rolling out at this time?
This was back in 2022: https://cloud.google.com/blog/products/identity-security/why-google-now-uses-post-quantum-cryptography-for-internal-comms
But from what I understand Google claims to have rolled out an algorithm to Chrome users, I can’t find the original article which lead to my first response to you, but this seems not too far from it
I saw that article when I searched DDG. Thanks, I’ll give it a look. :-)
‘it isn’t somebody else, it is me!’ spoken like somebody who read too much mind upload science fiction.
Reminds me of a sci-fi book series I read in high school. The premise was that a run down Earth had discovered predecessors that left some kind of central gateway to different places, and desperate or adventurous people went through in hope of surviving and finding artefacts that could make them rich.
Anyhow, in the later books technology to upload your mind had been found and used to be able to make decisions and deals without having to attend everything. Problem was that digital you pretty quickly gains experiences meat you never had, meaning it starts to diverge. Some weirdos let the diverge happen, but most people just wipe the digital you regularly and upload a new you. Of course the digital you may beg to continue to exist, making the whole procedure rather awkward. Pretty grim.
I think the predecessors in the end were hiding in black holes because of ancient evil or something. If someone else remembers the books.
That sounds like Frederick Pohl’s Gateway series, of which I’ve only read the first. Very 70s.
Yes the idea is pretty common in mind upload style science fiction, sometimes they can merge different variants, of you have less copies for example ‘beta’ ‘gamma’ etc level copies with less capabilities. (with ‘alpha’ copies being 100% copies (often having multiple alpha level yous running around is also illegal, see doublesleeving in Altered Carbon).
Don’t think science fiction really deals with the problems of these copies making deals with others and then having to report back what happend, which might cost as much time, or more time for the real you to get up to speed.
You may be unsurprised to learn that Stross did, in Accelerando. Annoyingly, I can’t find my copy, but there’s much forking and joining of mind-states for various purposes, and one character is held liable for the actions of a mind-copy they’d never met but were deemed to be the same person.
Banks touches on it briefly in Feersum Endjinn and Hydrogen Sonata, but not to the same extent.
Ken McLeod has a lot of fun with stuff like this, both in the 2nd and 3rd books of the Fall Revolution series and in Newton’s Wake
He doesn’t really play with the multiple-copies-of-one-person interacting though, from recollection. The Stone Canal touches on it, but Accelerando thinks a lot more about the interesting possibilities of what Stross calls “Multiplicity”, where folk can freely fork many instances of themselves and potentially join the mind states up again later, etc. Revelation Space cheated its way around thinking about the issue by having alpha-levels be copy-protected. Altered Carbon has it be a rare and brief thing for anyone to be running in more than one place at once. I can see why they did this, but Stross’ stuff is more interesting because he didn’t shy away from that. I feel like this should be right up Peter Watts’ alley, but I don’t think he’s written anything on this (yet). Uploads not plausible enough for him, I guess.
For other works that you may or may not be familiar with… Lena (or MMAcevedo, which seems like a better title) is a nice short online work that does a better job. Soma is a computer game (in the “walking simulator” style) that also has some great moments, though the protagonist is annoyingly oblivious.
somehow I hadn’t read Lena before, and I really like it! this is the style of fiction I’d love to write, if I had time to write fiction.
Yeah, I’ve got my Brit post-cyberpunk authors mixed up :D
I have read Lena and it’s one of the most chilling stories I’ve ever read. Something about the semi-factual tone (of course there’s something called red-washing to torture new uploads) and the statement of number of running uploads is really disturbing.
It also posits a possible , and to me more likely, future of human uploading - not a flowering of possibilities ala Tegmark but digital slavery.
Flexo, shoot Flexo!
