AI assistants are so back | The Vergecast
Вставка
- Опубліковано 1 чер 2024
- The Verge's Nilay Patel, Alex Cranz, and David Pierce discuss announcements from Google I/O and OpenAI's GPT4o event.
Further reading: www.theverge.com/2024/5/17/24...
00:00 - Intro
03:22 - OpenAI releases GPT4o
25:45 - Google IO
52:40 - Gemini products explained
1:01:56 - AI Overviews in Google Search
1:14:11 - Lightning Round
Subscribe: goo.gl/G5RXGs
Like The Verge on Facebook: goo.gl/2P1aGc
Follow on Twitter: goo.gl/XTWX61
Follow on Instagram: goo.gl/7ZeLvX
Follow on TikTok: bit.ly/42VeCVU
The Vergecast Podcast: pod.link/430333725
Decoder with Nilay Patel: apple.co/3v29nDc
More about our podcasts: www.theverge.com/podcasts
Read More: www.theverge.com
Community guidelines: bit.ly/2D0hlAv
Wallpapers from The Verge: bit.ly/2xQXYJr
Subscribe to The Vergecast on UA-cam, new episodes on Tuesday and Friday: bit.ly/3I6nJtz
If you buy something from a Verge link, Vox Media may receive a commission without exerting any influence on editorial content. For more information about our ethics policy, visit: www.theverge.com/ethics-state... - Наука та технологія
Vergecast has really grown on me.
See a doctor and you can get that removed
Y’all are genuinely funny & good at this. Many tech podcasts are bland or uncomfortably excited. This is great.
"Google is supposed to get you to the information, not be the arbiter of the information" Alex with the banger quotes as always.
And another: "The AIs don't think like humans do. They're not meant to. ... It doesn't think the way a human does, but we keep asking it to."
Always love when Nilay does the bit for the intro.
This is a weird thing to say, but I honestly *trust* Google way more than I trust Open AI at this point.
Google employees don’t even trust Google at this point. They used to have a motto of “Do no evil” and they got rid if it because it was getting in the way. Who should Caesar trust, the knife in the chest or the knife in the back?
Why?
I always trusted Google more but my trust in both is waning
This isn’t weird. It’s easier to trust the enemy you know and Google is well known more so than Open AI.
@@MikeAbili "enemy" may be harsh. Maybe that will change. But it just feels like Google is being much more careful about rolling things out than OpenAI. Again, this is just my personal feeling about it. Google has gotten much better over time, but mainly because it has been held accountable in the past.
This is how the technology improves. I do not understand the amount of time that is wasted harping on the point that it's not perfect
One thing I'm surprised you didn't comment about Apple is that their whole has always been vertical integration, but now they are licensing OpenAI for use in their OS and Google is the one that has the vertical integration with Android and Gemini.
Any vision based query is meant for glass. The use-case showed on phone and any or all use-cases were meant for the ML engineers or Data Scientists so they can find a use-case where it can be applied correctly. That’s why these are showed in a developer conference instead of a product launch day. That’s where I feel you are missing the point.
Also, seems Nilay missed (at 13:11) that in the Buckingham Palace example, the person using it was blind.
38:53 I think it answered the question fine as a person who has no knowledge of record players and had never heard the term before. I thought it made sense. If someone doesn't understand they can always ask follow-up questions. We do it in real life all the time.
Yeah this feels like a thing Nilay is a little too smug about lol. I love him but this gotcha isn’t as big a deal
“San Francisco is Milwaukee with rich people”. Nilay, you’re spot on
Love the donghuajinlong cap
Same, but _why_
@@impendio who doesn't love industrial-grade glycine?
Half an hour of tonearm discussion is peak Vergecast
I think Nilay's point about the record player answer is valid, but also kind of treats the situation as a void. We can assume the AI doesn't know anything about the user in this situation, so it has to just commit to *some* sort of level of detail for its explanation. So it commited to the point that the user knows what the counterbalance is. I think that's fine, because if the user doesn't know about it, they can just follow up with the question "what's the counterbalance?" and that just clears up the whole situation from every point of view.
Also I've seen a lot of media run with the "they've made AI flirty" angle. It feels like it's being brought up as the main story rather than the fact that they're just showcasing how emotive it can be. Sure, they could've showcased it with something else, but my point is that it should be "AI can be emotive" as the main story and "they showcased it by it being flirty" as the substory and not the other way around.
Is that a tax break in back of Nillay? The TV Frame he is always talking about??? It’s used on camera and now can be written off for work. Such a smart business man
Omg more appropriate B roll!!! Love you guys thank you thank you !!
RE: The rabbit spreadsheet demo
Totally agree that the use cases shown are often not really useful or at least not imaginative enough. A much better and definitely possible use case, I'd think, is showing the AI a whiteboard from a meeting where you've been writing ideas, dates, etc. and having it organize that info into a spreadsheet or other useful format. Or possibly showing it a receipt and having it automatically add the details to an ongoing spreadsheet.
The hallucination is that if we can be more efficient and clever consuming and playing with tech that the threat of global warming and global economic injustice won't matter.
The hallucination part of this whole discussion is so well observed and yet, it seems likely fixable? But maybe it is like self-driving and it will just get closer and closer and never really get solved. If that is the case, these tools are going to be WAY less useful than promised.
If you could fully trust what's available right now (say GPT-4 vanilla) it truly feels like it would kill any number of businesses but as it is, we seem to be a looooong ways away from that.
Another great ep.
Considering how much talk there was about the Decoder Podcast... Any chance of getting it uploaded to UA-cam again?
All ipad reviews for past 10 years: “Can it replace a laptop…now???”
What a brilliant episode, laughed a ton 😂 Thank you folks!
Thank you for the vid. Very enjoyable, I think the points tha Alex made regarding alien thinking are pertinent and will be plaguing the tech for a good while longer, which does mean ai will possibly be less useful in an everyday sense.
Welcome to the Verge Roast Show
Claude passed the river riddle just fine.
Younger generations are highly likely to embrace and utilize these technologies. My younger brothers and other family members almost exclusively talk to their devices, even in their current state.
A real test I do with most models, try it for yourself
'does blueberry have the letter t in it'
'no'
'but it does'
'sorry it does'
'where is the letter t in blueberry'
'it is the Xth letter'
25:00 OpenAI is basically the AGI Brotherhood of Steel
“I’ve been to San Francisco. It’s Milwaukee with rich people.” - Racine’s own Nilay Patel
It was a big week in AI? It was a week where the number of times people said AI was big.
Because five years downstream it is too prescient, we need to have a collective reread of The Big Nine: How the Tech Titans and Their Thinking Machines Could Warp Humanity by FTI’s Amy Webb
Amy has been on the Vergecast before! www.theverge.com/2020/3/31/21201103/amy-webb-futurist-predictions-interview-vergecast-podcast
@@reckless1280 thank you for that and this week’s discussion. A LOT of what you are cogitating about in this episode calls back to points she pinned in the big 9. I have not jumped into the Genesis Machine, but… 😳🫠.
Alex. Thanks for starting to do closure on your podcast opening stores. So much less anxiety I live with now
Visually impaired people could definitely benefit from that stuff if it works well and is fast enough
51:00 I don't understand the swing dance reference 🤔
the gold rush in niche news coverage now is: comic book movie and TV IP
29:00 - Cheap and fast until you corner the market, mono/oligopolise and then start increasing prices
Re: getting the right answer on the tone arm. Couldn't you just ask a second question on what the tone arm is? And presumably it would tell you?
Like I think it is really useful you might just need one follow up question.
I tried this trying to adjust the breaks on my bike. It was unable to give me a useful description or picture
I am once again begging you to get someone on who knows about emulators and isnt completely off base about 80% of this years old technology
Chat no longer gets the goat question wrong by the way. This is why it feels kinda futile to discuss the performance of a model in the first few days after 'release'. It gets tuned a little bit more during the roll out. There are so many logic questions it was getting wrong just 48 hrs ago that it now gets right.
Yes, general AI and self driving cars are one and the same. If one is reasonable, then the other is reasonable.
McDonald's time travel is wild
If you think abt it Moore’s Law actually covers “AI” advancement as well. Computer advancement is now called Artificial Intelligence. All these guys are building computers but they will rather call it something else.
And touch will always win. All these speaking stuff looks/sound good in movies and books.
@13:29 The guy is BLIND. The use case is for BLIND people. And if you can't see that, you are BLIND.
Our early appathy is going give them delusions of permission to design things that fit their vision of AI more than ours.
Like watching this on UA-cam music
If google does not send us to sites, how does google get paid?
I need my AI assistant to be so front! That being said; enabling Assistant to monitor all audio throughout the day, and commence recording extended telemetry data whenever specific words are detected (plus the addition of AI-assisted background noise removal, in order to isolate voices in loud environments like train stations) is exactly how my ARG-Evil-Pokemon-"Game"-app, aimed at business-people, is able to function so effectively. It's ridiculous how much real world, actual, genuine, information you can glean with selective enabling and disabling of Bluetooth radio while parsing gps data and using AI to comb for nearby posts on IG/FB/X etc.. to within 27metres of the phone, at this point I'm just typing for the sake of typing, and then Assistant will be able to tell you where to find apples.
Something i just learned just now is that Chat GPT 4o nows that the Verg podcast posted just now and It recommended to listen on apple podcast which is very interesting in my opinion why not listen on UA-cam why apple podcasts
Nilay is a glycine girlie?
Only industrial grade, please
0:58 All I can think about is the frame TV in the background 😂 I want one! The wife, not so much.
I think the problem with the latest versions of AI is that people still aren’t smart enough to know how to maximize their utility.
So true. The companies are obsessed with the arms race to AGI and have totally neglected to show this has any value.
Actually, I think it is that people are not smart enough to see the use cases. This discussion reminds me of the early days of smartphones. People routinely said "why would I need that, I can just use this other thing that has been around forever". The tech is legit. The hot takes are short sighted.
@@TerryBedford - I agree.
Kings cross station is most well known for a fire in which a lot of people died.
Make an iPad Pro OS that gives Mac OS abilities, and charge $100/year (or whatever) - i'm pretty sure people will eat it up
Latency.
First 😉
What‘s the point with americans wearing hats in closed rooms? It looks soooo funny! 😂
The fact that the auto-generated subs say ‘hello and welcome to rcast’ proves my point that Nilay speaks more and more sloppy and incomprehensible. Dude, articulate!
the lady on the pod seems jealous of gpt4o sky’s flirting
I don't understand what these journalist want from the LLM? Stop asking it facts? I just keep hearing hallucinations are the problem, but if that went away so would its creativity. I want the Verge team do define what they want out of one of these models, what use cases are they trying to solve for where its lying is getting in the way. Otherwise there's pools full of workers coding and creating who use them all day long and it just seems like they are stuck asking GPT-4 to report the news for them.
Talk to some people who use these as tools all day long. Like do some reporting.
We’re not the ones trying to replace Google Search with it!