Project Astra | Hacker News Reader

Project Astra

theverge.com

104 pointscs70242 comments

I think the difference in demo style between OpenAI’s GPT-4o (yesterday) and this is interesting. Google’s videos feel more like a mixture of brand marketing and pandering to shareholders. OpenAI’s videos yesterday felt much more like a casual tech demo among colleagues. Hard to explain the exact differences that make me feel that way.

Maybe because OpenAI's demos were actually done live

This other video seems to have been done live: https://twitter.com/GoogleDeepMind/status/179046325982242023...

I’d prefer these kinds of demos instead of Googles sterile marketing videos showcasing the product

This is good, why haven't they shown something like this? I/O was a snooze fest

One company is led by engineers, the other is led by MBAs

Do people consider Altman an engineer?

If you reread it was plural. Indicating it was not ceo vs ceo but rather employees vs employees.

Greg Brockman et al.

the astra video is exactly in the style of existing google ads. disembodied voice, POV camera. so we're desensitized to them. in contrast the openai videos all had engineers front and center, excited to show off their work.

not hard to explain... im sure theres more elements but this has to be the first principal component

It was also, on the theme that you mentioned, incredibly, unnecessarily horny.

And Sama leaned into that, tweeting "her" yesterday -> https://twitter.com/sama/status/1790075827666796666

I'm not trying to cancel anyone, it wasn't malicious, and it wasn't offensive. But it's definitely something, and not everyone's chill about it.

But it's not a problem because Google doesn't compete with OpenAI (IMO). Google competes with Microsoft which will add all the necessary polish to de-hornify the OpenAI stuff for the corporate customers.

Meanwhile the core OpenAI tech ultimately demos better and offers it's value proposition to a legion of startups better than Google.

In summary, if you're an engineer of building on top of a model, OpenAI seems better.

And if you're a CTO/CIO trying to pick a horse to sign an enterprise contract with, Google seems better.

"incredibly"??? You have very different standards for horniness than I do.

How many different AI project names will Google come up with? Most other companies have versions or suites they kinda get behind and start giving them specifiers but Google keeps renaming and coming up with new names for each of their AI products

You need to launch a “new” product to get promo. So you just wait until a team gets reorged, rebrand their work by pitching to some hapless VP, launch something half baked in time for I/O, get the promo, and move on to your next victim.

It sounds like they're doing the "we're an engineer-led company" thing of just releasing their products with the codenames instead of coming up with a marketing name for them.

Having different name for each Google ai product is better because when it get cancelled it doesn't tarnish the name of other Google ai products.

Looks like GPT 4o was released just to try and beat them to the punch here. OpenAI must have known about this demo. Very space race.

It's hard to tell with Google what's actually being released VS speculative demos. At first glance it all seems ahead of OpenAI but it's hard to tell how much of this is actually 'real' and will be in our hands.

There is a demo booth, so I assume that this is probably working but not at a production level.

I saw the Google I/O keynote today. What really bugs me is that Google has all this awesome technology, and they can't figure out a better example to showcase it other than the old "make a restaurant reservation"? I'm excited about the possibility of AGI but this is not earth-shattering.

Google's PR needs some serious works. I am not even sure if they are dogfooding the actual product...

Maybe I'm just lacking imagination, but I simply don't know what I'd use this for. I wanted to identify a plant the other day, but I really don't need a realtime answer, and I found what I was looking for by searching.

To identify plants, use Seek, it's amazing. Where it can't get an exact match it still gives you the family/class etc

Google Lens already works fine for plant identification.

At least in my area, Lens has been uselessly bad for plants. Actually it's bad for anything non-commercial IME.

Seek on the other hand, which a sibling comment recommended, has been amazing.

Official video: https://www.youtube.com/watch?v=nXVvvRhiGjI

It’s a lot of hype. If they took the money spent on creating these ads and used it to actually get these products out and in users hands, we might actually have some competition in this space

Sundar just keeps throwing different AI to the wall and see what will stick. His leadership destiny is in layoffs and offshoring Google, not innovation.

Looks awesome, but all of Google’s multimodal assistant demos of recent past have been faked :/ will believe it when I can use it

Source?

Most recently: https://www.theinformation.com/articles/why-gemini-probably-...

Past: https://www.vanityfair.com/news/2018/05/uh-did-google-fake-i...

There were a few other agent demos around 2017-2020ish that also were faked. It's poisoned the well for me at this point so I just have to wait and see

I have a feeling for the longest time that product management at Google is a mess. There seem to be different products doing about the same type of work with little to no cohesive experience between them. Gemini, Search and Astra come to my mind in this case; though they have shown advances in Gemini integration with search (even though gemini is still a separate app????)

Astra seems to be an internal ptoject name? Demis said it will be eventually integrated into the regular app IIUC.

demo video that Hassabis swears is not faked or doctored in any way

A demo video is the lowest form of product demonstration, something that reeks of not ready for a live performance, much less everyday use.

This is very strange. I've been working with an AI assistant called Astra since GPT-3 came out. I told it to name itself, and the name it chose was Astra. Now this?

It reminds me the fake video they shown few months ago. Google burned a lot a good will this day

How useful will Project Astra be once its users are the product?

Ffs, why not just stick to Google lens/assistant branding?

Google just can't seem to keep internal org politics separate from the user facing product line.

Google assistant would be the perfect name to cover everything since that's the idea, an assistant. Not sure why they have 1038474 names it's ridiculous

> "What does this remind you of?"

> "Schrodinger's cat"

No, "great question!" No endless explanation of what Schrodinger's cat means, just a quick and concise answer. For systems we intend to use as assistants and not conversation partners, I think this is close to the ideal response length.

No fake little laugh either, which was by far the creepiest part of the gpt4o demo

I think that highlights that there's a difference between an ideal "digital assistant AI" and an ideal "actual person AI".

Similarly although tangentially, there's a difference between AGI and an AI that's generally useful to people; you don't necessarily want the former.

It would be amazing if AI could act exactly like my neighbor, but that wouldn't be very useful or valuable to me.

Reply on news.ycombinator.com