@chiasm
There are no details I have seen. But my expectation is that the car - detecting incomplete data- stops until it receives confirmation from a human that it can continue.
@garymarcus
Tagged: reply
@Zeugs
You may be interested in my post from a few weeks ago:
Benefits and risks of synthetic video and audio
https://axbom.com/synthetic-video-benefits-risks/
@Zeugs
Not sure I follow.
I've mentioned the effects on my own interpretation and emotion, which can happen regardless if others would agree that it represents me, or regardless of whether the output can be deemed "good" according to some definition of that metric. Outer representation is a tricky subject. I mean, people can be led to believe that an empty rubber glove is their own hand.
Perhaps if you let me know your goal with the line of questioning, or your own assumptions/hypothesis/conclusions, this would go easier. š
> why are you suspecting that GenAI video's are good?
Did I express this within the context of this conversation? Not sure what "good" refers to here either?
@Zeugs
> how good are you in judging your Own outer representation?
I wouldn't know how to judge good or bad in that respect.
> Could you judge if it's good Korean?
Nope (but I am one month into studying Korean with Duolingo. š
)
> Does a voice identifier, identify you as the speaker of the generated Korean?
That's an interesting one. I'll have to check with some different ones. It wouldn't surprise me if the answer to that is different for different identifiers.
@Zeugs This is a great philosophical question. What Iāve used here is the basic version of something that can be more powerful for a hundred bucks. Sure I think I could identify these videos as fake if someone else made them, but it all depends on context.
I do think weāll get to a point where people will start questioning themselves and their own memory. But even if I can see itās fake, what happens when a majority of other people thinks it isnāt? Can they gaslight me into accepting that itās real?
But already I feel at odds with what Iām seeing. I can even catch myself thinking, āWow, my Korean pronunciation is really good!ā even though I logically know Iām not voicing those words at all.
@kefir Sure, this is all HeyGen. I've shared some more here:
https://axbom.com/synthetic-video-benefits-risks/
It's possible to pay to get better quality.
@salixsericea Thanks for the feedback. Yeah, I really agree about how it looks like a dubbed movie. Apparently customers of HeyGen can pay to get better lip sync. Of courseā¦
Makes me think of the animated film Robots and hunting for upgrades.
https://www.youtube.com/watch?v=99DeRHisQtI
@Zeugs Not that I watch them that regularly, but across the years many of my talks have been recorded and posted online, so it happens. I do regularly listen to myself, since I run a couple of podcasts and do production for them.
@onepict I've been toying with the idea of giving someone control over "me" for a few weeks. But this also feels terrifying. And that's coming from a middle-aged privileged white dude.
Probably why I should do it.
@onepict We have a situation now in Sweden where the leader of the far-right party is about to publish a video (he has announced) where he is speaking Arabic and telling people they are not welcome hereā¦
@onepict They're also AI-generated. Including voices. Except one, where I uploaded my actual Swedish voice.
You can see the whole showcase here:
https://vimeo.com/showcase/10723290
@onepict Agree about the uncanny valley. I see a lot of risks amongst the benefits.
https://axbom.com/synthetic-video-benefits-risks/
@Mabande Good reference!
@ojala That's a really good and relevant observation. Thanks.
@clive @mollyanglin Thanks Clive! It's funny, a friend of mine pointed this out exactly about email when we talked about it yesterday. For many of us it has left a mark.
@lisamelton Wow, I'm so glad to be able to read a story like this in these times. Thank you for sharing.
@mollyanglin @clive Wow, that was bleak. But good writing. Thanks for sharing!
@ianbicking Not my intent.
I would love for these voices to talk more about the value they are creating and how they are receiving feedback from the people they are hoping to help. What I'm questioning is the idea that working faster necessarily means more value created.
The work being done IS often valuable, and we risk losing sight of that value when speed becomes the prioritised metric.
I do my best work not when I work faster but when I spend time truly understanding the context and expectations of the people I hope to help.
The factory analogy can be helpful when considering how me suddenly working faster can affect the next person who has to manage my output.
@rodet Definitely. I was just discussing backlogs in another forum around this topic. And this works if we are certain the more āefficient codingā doesnāt create other issues by way of documentation and traceability.
Of course, if the documentation and traceability was never there to begin with there is not much of a difference. š
@autiomaa Good example! I really can work so much faster if I do a poor job.
Meanwhile, if I work slower my quality generally improves, I make better decisions and I save others a lot of time.
@autiomaa Exactly. In an organisation that works in silos one department can increase their output, be praised, win awards, receive more funding and the employees get raises. Meanwhile the department that has to deal with the sudden decrease in quality has a much harder time getting their work done, and suffer the consequences for it.
@hbuchel Wow, I feel this so much it hurts. Thank you for writing it!