The New Bard and AI Images, Videos, and Translations

The new Bard is actually useful, even if you look past the hype. I'll show you why, and cover how you can make the AI images that are going crazy viral. It's genuinely fun, and surprisingly customisable. If that wasn't enough, you'll learn about GPT 5 red-teaming, HeyGen translations, Alpha Missense, and much more!

[AnandViswanathan]

The trippy thing about that Hindi dub is that it uses some English words in exactly the same way that an Indian person would in casual conversation! Amazing!

[geckomaniac]

I agree with you that AGI will be more than just a chatbot. It will have the ability to solve complex problems and make creative decisions in a way that is currently beyond human capabilities.

[janniksco]

0:31 Bard Extensions
5:46 AlphaMissense Mention
6:08 HeyGen Avatar 2.0
9:31 Metaculus Ad
9:58 Guests
10:18 AI Image Generation
12:22 OpenAI Red Teaming

[wesremy]

13:05 the longer i look at this picture the more impresive it is. The way it uses pillars, shadows and trees to create letters but still creating a logical image is just amazing

[false]

I appreciate your focus on the ethical implications of AGI. It is important to think carefully about how we develop and deploy this powerful technology

[ben_sch]

Your video was well-argued and informative. I learned a lot about the potential of AGI and the challenges we face in developing it responsibly.

[perplexedon]

Oh my goodness, that AI Explained image is so subtle and clever. That's the kind of creativity I previously assumed was still a little bit away. I had to pull the phone away from my face to see "Explained", but the detail of the buildings fully holds up on fine inspection. Crazy

[VedantinKK]

Thanks for being the best AI tech communicator out there dude.

[Lishtenbird]

I have just a faint glimmer of hope that a recommendation engine like this might shift priority for creators from endless shocked-face clickbait, to actual meaningful and informative content.

[svka]

I'm an engineer and I asked Bard to give me the proper size for an A/C system for a house. I inputted all the usual info I would enter into our sizing software and it gave an accurate answer. I was blown away.

[Citrusautomaton]

The AI dubbing is out of this world! I recently watched Iron man for the first time and in it there’s a scene where Pepper uses a Stark computer to “auto-translate” what the Ten-rings were saying. When i first saw it i was like “yeah, that’s not that far-fetched”.

[DaveShap]

"I love how we get to see some characters randomly yeet out of the image"

[haileycollet]

If they're ramping up their red teaming network that can only mean one thing: they'll have something that needs red teaming soon. They tested gpt-4 for six months so ... soon plus something like six months....

Also, would be nice if the demo translation videos included some TOO English. I sorta speak enough Spanish to get a feel for the quality (pretty good!), and I mean you can just try to see if the sounds line up with lip movement, but it's hard to tell if it really feels natural

[Anders]

The AI dubbing is brilliant. In the near future I imagine that this functionality can be added to smartglasses and then when someone speaks in a language we don't understand the smartglasses can dub the person in real-time.

[ThalanorThornhale]

I wanted to make a comment about hallucination:

Heavy epilepsy patients undergo hemispherectomy which is a surgical procedure that separates left and right brain. The following experiments were done:

When an image is shown to the right eye (e.g.: a saw), and simultaneously another image (e.g.:hammer) is shown to the left eye, the following happens:

When asked what the person had seen, he will respond with the word "saw". When asking the person to draw the image with his right hand, he will draw the hammer illustrating independent right, left brain function.

Here is the kicker that I had to think about when you mentioned hallucinations: when asking the person why he drew a hammer instead of the saw, the person said that that image was probably something he saw on his way in!

2 inferences I draw from here:

1.) Human brains can "hallucinate", too!
2.) Hallucination may be caused by lack of context. In the person's case, it is the lack of knowledge that the other side of the brain caused the image to be drawn.

Within the computational context it may be that we don't have other networks producing corroborating context data (sounds, touch, visuals) and the resulting emergent property of self-awareness.

We are only feeding text.

Maybe we need independent networks for visual, sound, touch and smell, and then integration networks that may contain LLM architectures.

[adityashukla]

You're the only channel I like to watch to get my dose of AI.

[YUTPIA]

>we're putting together a team to redteam our effort to make AI ethical and aligned with human values
>philosophy? what's that, some kind of exotic mushroom?
amazin'

[jimbetts]

I don't think AI is hallucinating. I think AI is imagining. This imagining will ultimately lead to singularity.

[TheBitcoinBoomer]

I have dozens of PDF service manuals in my Google drive, Bard was able to accurately pull up part numbers with only the PDF name. I tested it about 15 times with 100% accuracy.

This is a huge time saver that I will use often, but haven't really found any use case for any of the other extensions.

I was hoping to get video summaries from YouTube, but no luck.

Thanks for the video!

[BrianGlaze]

I've been using Bard quite a bit and I like the upgrades over time