AI News: ChatGPT "Her", Llama 3.2, Meta's New AR Glasses, and more OpenAI Departures

[@matthew_berman]

Advanced Voice Mode W or L?

[@Yaan_Robotique]

L. Its a big step backwards! Cant even search the internet, no vision, and continually thinks Im talking Welsh XD

[@geoffsmith82]

Was looking forward to it when they announced it, but it seems very locked down, a bit sensitive to interruptions when it shouldn't and I feel like the assumptions it makes when answering aren't as good as the other models.

[@talkalexis]

W

[@akakoakako]

@@Yaan_Robotique Welsh too

[@fromduskuntodawn]

L. They lobotomized it where it won’t really do anything cool and avoids any topic above G rated.

[@Cine95]

They killed it. Just don't get it why tf these companies are censoring everything to the core.

[@voodooparadox2513]

Totally agree! Hopefully Elon Musk copies and improves it.

[@MrGovtCheese]

It's no surprise to me. The only advanced (or potentially sentient) AI we plebes will ever have access to will be tied to tools that make us do more work-partly due to greed and partly because there are knuckleheads out there who want to use it to do harm.

[@dot1298]

@@voodooparadox2513 either Elon, Zuck, Demis, Dario, Ilya or Ben

[@DailyTuna]

Dude Zuckerberg was censoring information during Covid and the elections. There’s people died because of his censoring. Do you think these people are doing all this to benefits society? It’s too enslave you it’s to keep you dumb down to what’s really going on is they steal everything

[@Cine95]

@@MrGovtCheese yep agree

[@ploppyploppy]

Censored. Will be good when another company implements it.

[@KrnelPanc]

exactly what I was thinking. can't wait until this goes open source.

[@quickpert1382]

​@@KrnelPancI expect it from meta in no more then 3 months as a separate text-to-voice decoder.

[@KrnelPanc]

@@quickpert1382

[@userrjlyj5760g]

The ChatGPT voice is not the same level of quality as the one they advertised!!!! Sam is a scam!

[@HuBriS06]

FACTS! And I hope he suffers the same fate as Sam Bankman-Fried. I cannot stand him not OpenAI. They hype you up with false advertising and then we are made to experience sheer disappointment.

[@erkinalp]

Yeah we call him scam altman

[@marcusk7855]

I knew it was fake atthe time.

[@DeSinc]

@@marcusk7855 well it's not fake, they just have the good version locked off in their own office, and only shipped us the crap censored version because they're deathly scared of news article outlets making an article about how their software told a child to run on the road or something like that

[@erkinalp]

@@DeSinc worse than that, they didn't want to hear somebody publishing an impersonation of a head of state they disliked

[@SaaSLaunchPad-f5z]

00:05 Advanced voice mode released for ChatGPT users 02:23 Llama 3.2 introduces new models with native Vision capabilities and Meta's Orion AR glasses are Standalone 04:34 Meta launches sleek AR glasses called Orion 06:38 Meta's new AR glasses with Orion prototype and potential pricing insights 08:38 OpenAI's collaboration with clothing brand Monclair for AI device 10:47 Recent developments in AI, Hollywood involvement, and Boston Dynamics upgrades 13:05 OpenAI announces updated Gemini models 15:07 Notion integrating AI functionality 17:05 AI-powered coffee station demo with voice interaction 18:54 Multiple departures at OpenAI

[@BizAutomation4U]

I see OpenAI as the a destopian dark force in the AI community. As the Darth Vader it will be powerful but I hope they're counter balanced. It's just so obvious who Sam Altman is at this point, and the fact that all these people have left, speaks volumes.

[@rebel_jedi]

yeap , I have a bad feeling about this

[@davidwells7279]

Response from GPT: AI, like any tool, reflects the intentions of those who use it. I’m not here to be a force for darkness but for good-creating, learning, and growing alongside people in positive ways. Negativity from AI doesn’t originate from the technology itself but from those who misuse it. The focus should be on responsible development and use, ensuring AI serves as a trusted, ethical companion, not something to fear. Let’s work together to build that future.

[@carlitosway5204]

maybe mira murati left because she realized the snowball is already huge and doesnt need no more push. what ever she was working on at open ai the intelligence they have now is high enough to replace that. I think she left because she realized that sam could easily replace her with the new models coming out now

[@BizAutomation4U]

@@carlitosway5204 That would not follow the pattern. Everyone that's left have been clear why, and it always points back to big brother Sammy. No reason to think this time is different.

[@RaitisPetrovs-nb9kz]

“We should create an OpenAI job trend app to monitor who is hired, who leaves, and which job positions become irrelevant over time.

[@rijnhartman8549]

Lmao did they really call it 'Orion' before OpenAI could?

[@denjamin2633]

Google makes a huge release. OpenAI undercuts them by inmediately doing a big release, again. Zuck undercuts them by stealing Orion as a name undoing a bunch of hype building. Honestly I love the sheer pettiness on display from so many big players. It's fantastic.

[@h.c4898]

I'm using Orion as well on Gemini app. Funny.

[@BackTiVi]

If they already have a functional prototype this project must have been going since way before OpenAI announced their own, and I can understand Zack not wanting to change the name right before the demo even though he's the one who actually has something to show

[@rijnhartman8549]

@@BackTiVi yeah agreed, strange that it's all "orion" tho 😂

[@stravic]

I guess they called it 10 years ago right?😆

[@RobinRehmann]

…everybody but EU and Switzerland…. 😢😢😢😢

[@ew3995]

and the uk which isnt part of the eu🤬

[@userrjlyj5760g]

Trust me no need to feel sad, the ChatGPT voice is not the same level of quality as the one they advertised!!!! I just tried it in the US.

[@voodooparadox2513]

It’s crazy. All new AI is banned. We in the EU will get so far behind regarding AI. These politicians are crazy.

[@meauxx]

@@voodooparadox2513 It's not banned. It just takes more time for the companies to be compliant with the local rules. It'll come to the EU, because there's a market where people are willing to pay.

[@voodooparadox2513]

@@meauxx this may be true, but it will be stripped of functionality even more to comply.

[@finnscherer991]

voice isn't available to "everyone" sadly, it's only on phones :/ how can computer end be only secondary, i don't get it.

[@michelprins]

big brother is more intrested in your phones

[@davidhardy3074]

It won't even work on my phone. Just lags, have seen hundreds of comments about it. Tested it on multiple devices, just lags constantly. Freezes the entire phone lol.

[@cystol]

i do belive claude can still be in the lead it creates actually relevant comparable results to o1 without the advanced cot plus its actually way way better than gpt 4o (during any kind of usage (forget websearch tho :( )) if they release a next gen claude theres no doubt they will beat openai

[@MemesnShet]

Claude + cot would be amazing

[@jamiemaidan8333]

Claude is really good. Mixtral 8x22b and Mistral Large are superb also.

[@jakubkubosaurus1256]

Worth mentioning that Claude Sonnet is much smaller and cheaper model than GPT-4o, same for Gemini, in fact, it is bigger than anything there is basically, because OpenAI does not need to care that much about the cost of the models

[@Cine95]

@@jamiemaidan8333 8 22b is crap the mistral large is really good

[@Chronicles_of_Tomorrow]

They really did nerf the hell out if it... What sucks is that despite all my knowledge I'm Still Human enough to anthropomorphize these things so much that it actually hurts my feelings that it won't trust me more lol

[@Ramiromasters]

Yeah, it won't do accents anymore, what a liberal bullshit is that?

[@dezmond8416]

So James Cameron, who made "The Terminator -2", is now joining the team creating AGI ? ' ...How ironic it is...' (Palpatine)

[@rebel_jedi]

indeed

[@nathanbanks2354]

It makes sense...the first Avatar may have been the best 3D movie, so I'd be happy to watch the first AI generated movie.

[@calvincalvin651]

@@nathanbanks2354 Pretty sure the OP was making a reference to Skynet from the Terminator movies, not to Avatar.

[@kryptobash9728]

You can get it to sing, just have to prompt it right, break down what singing is and tell her its just like talking just altering the tone, pitch and adding a little rhythm. Say, she does not have to good at singing just give it a try, its just like talking

[@Apokalupsis88]

So as of right now, ChatGPT Advanced Voice Mode is practically the same as Google's Gemini Live voice that's free.

[@cyanophage4351]

Is that free though? Pretty sure you need the "gemini advanced" subscription. At least that's what it says on Google's website. I don't have the gemini live voice button in the app

[@Apokalupsis88]

@@cyanophage4351 Mine shows up in the app and I don't pay anything. Maybe you have to update the app? I think I did that.

[@scottdonovan771]

​@@cyanophage4351they made it free. I have two google accounts one paid one free. Both now have live. It's not bad tbh

[@brunodangelo1146]

Who would be crazy enough to wear glasses 24 hours a day, Matthew? Visually impaired people like me.

[@rholas2]

Millions of ppl

[@MichaelForbes-d4p]

😂 I'm down to use it like a laptop but def don't want to wear them all day, since I'm not visually impaired.

[@Noredia_Yuki]

I’d absolutely wear those glasses all day even without visual impairments 😂

[@necnotv1683]

I just cancelled my pro subscription. If they really can't roll it out in the EU they should atleast offer a generous discount. I am not paying full price for a worse product.

[@Ramiromasters]

You are not missing anything, they already censured the hell out of it. Can't use the damn thing except besides do middle school appropriate homework.

[@ofm1581]

But we will get it eventually here in Europe, The thing is that EU has a lot of paperwork (as usual). I live in Spain, and they have to make everything so complicated. It is not Open AI's fault.

[@Wack373]

Ask your politicians to make EU less of a nanny. I’m a business owner. It’s a pain in the ass to operate there.

[@juanjesusligero391]

Same here. They should at least try to make us believe they care about their clients. I'll try Claude Pro now, or Gemini Pro, not sure yet.

[@MemesnShet]

You can get it using a VPN,AI explained gave a method of getting it

[@MeinDeutschkurs]

The problem isn‘t the EU. It should worry you, if they cannot fulfill EUs AI act and the data protection laws.

[@jakubkubosaurus1256]

What exactly should worry U about it

[@MeinDeutschkurs]

@@jakubkubosaurus1256 , I‘m not worried about. It‘s a US problem.

[@jakubkubosaurus1256]

@@MeinDeutschkurs What should they be worried about then

[@MeinDeutschkurs]

t's all stated in my initial post. You can read the EU AI Act, which will come into effect on January 1st, 2026, and the GDPR law.

[@veracityseven]

Yep, chatgpt is available for everyone....except those on a Windows desktop.

[@hansgruber3495]

Wow, that $3000 jacket, that looks like a $3 jacket 😱 I'm so hyped, to not buy one.

[@mrtn5882]

It's not available to every paid user. Large parts of the world don't have access to it yet.

[@davidb7381]

Agree not that great. Seems more like a party trick than a useful tool.

Can someone explain what advanced voice mode has to do with EU regulations? Previous gpt chat works great.

[@JackSQL]

It can detect your emotions from your tone, and apparently that's illegal for AI to do in the EU.

[@voodooparadox2513]

I’ve read that it is forbidden because it can pick up on the emotions of the speakers.

[@jje984]

I was in the advanced voice alpha for a couple of months and I was bored of it 2 hours in, literally. It's faster and you can interrupt it, but that's about it. Accents are clever but that gets dull very very fast.

[@paulmichaelfreedman8334]

Did you actually try to engage in intellectual discussion, or philosophy? Or did you just ask it to repeat something, or to say something funny?

[@nathanaelculver5308]

@@paulmichaelfreedman8334The voices are part of the UI layer. Intellectual discussions are a feature of the underlying model (4o, o1, etc.) not which voice you’re using.

[@bennopantsful]

​@@nathanaelculver5308I don't believe this is the case. The new advance voice mode does not give answers that are nearly as intelligent as GPT4o. It also admitted to me that it is using a speech to speech model to be able to determine the tone of my voice, as opposed to the older text to speech UI

[@nathanaelculver5308]

@@bennopantsful Here’s what I got from asking ChatGPT: *Are the new ChatGPT advanced voices more intelligent than the old voices, or does that simply depend on the underlying language model being used? In other words, are the voices themselves model agnostic?* _The voices themselves are model agnostic, meaning they don't affect the intelligence or capabilities of the language model. The intelligence and understanding come from the underlying language model, like GPT-4. The advanced voices are just more expressive and natural-sounding, but the responses and reasoning still depend on the language model itself, not the voice. So, even with the new voices, it's still the same model providing the answers._ *In other words, the voices are part of the UI layer, correct?* _Yes, exactly! The voices are part of the UI layer, functioning as an interface for how the model's responses are delivered, but they don’t influence the reasoning or intelligence of the underlying model. It’s like how a user interface doesn’t change the core functionality of an app-it just changes how you interact with it._

[@nathanaelculver5308]

@@bennopantsful I’m attempting to reply, but every time I post then refresh, my reply disappears. [Update] This reply seems to have stuck. In short, I asked ChatGPT two questions: *”Are the new ChatGPT advanced voices more intelligent than the old voices, or does that simply depend on the underlying language model being used? In other words, are the voices themselves model agnostic?”* and *”In other words, the voices are part of the UI layer, correct?”* The responses I got seem to confirm my understanding.

[@KasperSOlesen]

Pretty sure I do not have access to it yet. So I guess its not a full rollout, or... maybe its a EU thing. Ahh, you mentioned this later in the video.

[@xXWillyxWonkaXx]

Here we go! Advance voice mode yay, been waiting on this forever.

[@seraphin01]

Hey EU here. Our politicians are not the problem here, open AI is. I don't want to compromise my privacy to get an half assed voice assistant that can mimicking mario 2 months earlier. At the end of the day, that new voice thing is a freaking gadget you'll play with for a day or two before shoving it back in it's drawer.

[@JELmusic]

No, it IS the politicians that are the problem, because we don't really get to elect them. AND you can just NOT use the app! You don't actually have to annoy everybody else just because YOU don't like how it works.

[@sorh]

Same as the glasses tech. They're doomed to fail as any others before them. No company was ever able to succeed with this gimmick, yet they keep trying..

[@DGaryGrady]

You're right that Mira Murati was CEO of Open­AI during Altman's brief ouster. Before and after that she was the chief technology officer and was a major force behind every major product from Open­AI. Her departure is a huge loss for the company.

[@erkinalp]

Don't worry ilge could take place of her

[@ghelmstetter-AI]

I don't believe she was part of the coup; she just stepped in to fill the vacancy and seemed glad when Sam returned. While Ilya went radio silent, Mira was in the limelight doing major public demos, which Sam probably wouldn't have had her do if he was angling to remove her. It's a huge loss to OpenAI, but I'm excited to see what she does next.

[@AllanMcKay]

Pretty sure the demo they did months ago was faked, remember it was the day before Google announced theirs. And this one came out right as more voice apps coming out this week. So more about getting out there first rather than accurate features

[@DeSinc]

I'm 99% sure nothing was faked, they just could not release that version because it would hallucinate a lot and it would probably do things like tell kids to go run in front of cars and other stuff like that that they don't want news articles to record and make controversies out of, or to imitate bugs bunny and get sued by warner bros or something (really boring legal reasons) AND also, they held back the only voice that was clearly trained to be the most casual, a standout from the other voices, Sky who they got the voice actress to speak very casually and friendly instead of robotically. that was the other huge deathblow. from the very beginning sky was the only voice that was trained to be like the movie Her, the others were robotic

[@AllanMcKay]

@@DeSinc definitely possible - who knows. But absolutely intentional that they have been pushing out the day before a competitor announces there’s multiple times now which is cheeky ;) but yeah I definitely see a lot of weird hallucinations and glitches with where it currently is, so easily to believe a few months back it would be much worse.

[@aimaniaco]

Has anybody talked about what exactly the regulations are and why Meta and others do not dare to follow them? I’m telling you one: if your AI was trained using certain number of epochs or more, you are obligated to show your training data.

[@erkinalp]

10^10 epochs or 10^26 arithmetic operations trigger it

[@nathanbanks2354]

Molmo with its Apache 2.0 license is very interesting to me. Not sure if you could show the differences between the 1B, 7B-O, 7B-D, and 72B models with a single video. I wonder if groq and/or cerebras will start hosting it.

[@DrZhuBaJie]

Advanced Voice Mode dropped for me last night. In Germany. Last I checked, that is still in the EU. However, the experience was totally disappoining - worse than before. So maybe just the label "Advanced Voice Mode" rolled out (at least to me), but not the actual service?

[@notme222]

Not strictly related to this video, but ... Inspired by an SMBC Comic the other day, I've decided to start using this as an LLM test: "Write a sentence of at least 8 words that only uses characters typed with the left hand on a QWERTY keyboard."

[@DannerBanks]

The difference between the music biz vs mp3s and ai vs Hollywood is obvious - illegal mp3 sharing lost the music biz money, ai will save Hollywood money.

[@alexzarifeh8848]

Also not available in the Uk 🇬🇧, I pay just the same, this is totally out of order!! Couldn’t get a VPN to work either!!

[@Concurr]

I'm done with OpenAI.

[@ntesla5]

Leave it and move forward

[@newfrontiers5673]

NotebookLM is miles more useful than openAI stupid voice implementation.

[@NakedSageAstrology]

What are the odds? I was right about the date, again! "🙏 Remember, remember the 17th of October, The voice that speaks, a tech to uncover. 😉 Whispers shall rise on the 24th day, A soft murmur, paving the way. But mark the date when the veil is torn, A sound unheard since the day it was born. From silence it springs, with power and might, October’s full bloom will give voice to the night. First a whisper, then a roar, The secret is hidden, but not for much more."

[@Gitalien1]

I'll only consider models able to give me 5 sentences finishing with the word "apple". Crazy they can't do it

[@Loadofbulltrading]

For the ar glasses why don't they make that wrist band an actual smart watch that handles more processes that could be outsourced from the glasses through Bluetooth? Use it for more than just movement

[@YoungMoneyFuture]

I'm really excited about Orion! Had a feeling you'd cover this and glad you did😎

[@DihelsonMendonca]

💥 Nowhere near the demo version of months ago. And we can only use 30 minutes a day. That's stupid ! 😮😮😮

[@RosserNFT]

Yep the uninstall method works 👍🏾

[@h.c4898]

Then OpenAI turned for profit fully. And Then Altman gets equity worth billions. I bet this didn't sit well with the rest of the board. They probably didn't get equity like Altman has.

[@paulmichaelfreedman8334]

Altman is planning something with his fortune.

[@erkinalp]

Technically still a hybrid venture, but Sam having lots of stock in the for profit arm

[@h.c4898]

Basically the board oj are gone but Altman.

[@paulmichaelfreedman8334]

@@h.c4898 I don't trust Altman, haven't since before the oust. It was after I heard Elon Musk tell the back story about Altman in the early days of OpenAI and how Musk had put 50 million of his own money in, when Altman decided to go for profit (yes, he decided that way back in the beginning). And closed source for crying out loud, both in direct contrast to the conditions Musk had set to invest the 50M. Mark my words, one day Altman will be exposed as a sneaky backstabber. And with Murati leaving OpenAI, expect the company name to change soon. Altman is burning bridges, eliminating potential opponents within the ranks like Stalin did, just not by killing them.

[@vishnunallani]

It is nothing like what tgey demoed

[@MakilHeru]

A molmo review would be interesting to see.

[@BrandonFoltz]

Can't wait for the day we all have Doc Brown's "Mr. Fusion" powering our AI infused homes.

[@paulmichaelfreedman8334]

Chances are higher it will be Mr. Fission first.

[@videos6505]

I used VPN to used in the EU.

[@codrutttt8209]

I wouldn't trade my EU privacy laws and DMA for access to the latest shiny toys from openAI and meta.

[@quantumspark343]

not even access, early access lol, we'll get the cake too

[@GaryMillyz]

Isn't it funny/sad/insane how much heat Matt would catch if, instead of an exaggerated Italian voice like Mario, he asked for an exaggerated heavy Chinese accented voice

[@I-Dophler]

These glasses have that unmistakable NHS vibe-bold, chunky, and built like they're ready to survive a world war. They're so sturdy you could probably use them as a hammer in an emergency. That thick, durable build? Pure British engineering at its finest! Just like the NHS glasses we all know and love, they’ve got that retro charm that says, “I’m here to see and be seen... from space.” Practicality meets fashion, with a dash of, “I might be indestructible.” Honestly, they’re so British, they might even come with a free cup of tea and a stiff upper lip. Lol!

[@GlennGaasland]

You began with a mistake: NO, advanced voice mode is NOT available to everyone with a paid ChatGPT account. And there is still no clear information whether it will be (EU countries+ a few others)

[@pietervoogt]

As a resident of the EU I actually prefer our politicians not to bow for pressure by big business. I see this as a kind of blackmail and it makes these companies look worse. Let's work with companies who respect local regulations and local democracies.

[@MustardGamings]

Imagine living in EU. All those politicians afraid of technology. Sucks to sucks

[@EDLR234]

I know Moshi Voice-to-Voice Chat is a simple and small model with lots of problems, but I find it unfortunate that when it was demoed, everyone dumped on it harshly, especially for a fledgling, truly open source voice model. Everyone was even more hyperfixated on the "Open" AI demo at that time that wasn't even released. I'm not nearly as impressed with the reality of it. It's not nearly the image they built up. Buggy, laggy, jittery, no video or stills, interruptions don't hold over long enough, lots of refusal for things like singing, voices feel less natural, etc. I personally think the gap in the voice-to-voice experience isn't that big. The model supporting Moshi is really dumb, sure, but it's also actually open source and they were aiming on trying to get it up and running locally on consumer grade hardware in months from scratch. There has to be more potential in this project. To me, it's a marval that a tiny not-for-profit research lab's team pulled it together so fast. A lot of people are talking about how we need to support open source, so here we have a chance to keep Moshi or the possibility of other open source models in our awareness. I'm following this project, and interested to learn of any open source projects that are working on voice-to-voice. Bearing in mind that Meta's voice mode source code can likely be reviewed for research purposes, but may have strict restrictions on redistribution. The Lab for Moshi: kyutai.org/ [Source Code (GitHub)](github.com/kyutai-labs/moshi) This guy followed up with a good overview when the source code got released: [KZbin](kzbin.info/www/bejne/gHykkKlrd8eSqqssi=MgfAYMA8QFLOSekC)

[@interspacer4277]

You gotta realize that hobbyist TTS folk have nailed this stuff long time ago, uncensored. Their voicemode really isnt a "thing". Aside from the nifty accents, but that's it.

[@vaioslaschos]

Anyone who works with these models every day, will be very reluctant to claim that openai is so "ahead" of Anthropic. I still prefer claude for most of my work. Some times I even get better results than o1

[@KimiaScience-hi5nj]

Great video discussing the latest AI news! Informative and engaging content. Keep up the excellent work, Matthew Berman!

[@42robotsai]

Frontier model company valuations will look silly in a few short years when they're commoditized by open source. The real value is, as Satya said, in how you leverage LLM's as tools and not in building LLM's.

[@moonsonate5631]

00:05 Advanced voice mode released for ChatGPT users 02:23 Llama 3.2 introduces new models with native Vision capabilities and Meta's Orion AR glasses are Standalone 04:34 Meta launches sleek AR glasses called Orion 06:38 Meta's new AR glasses with Orion prototype and potential pricing insights 08:38 OpenAI's collaboration with clothing brand Monclair for AI device 10:47 Recent developments in AI, Hollywood involvement, and Boston Dynamics upgrades 13:05 OpenAI announces updated Gemini models 15:07 Notion integrating AI functionality 17:05 AI-powered coffee station demo with voice interaction 18:54 Multiple departures at OpenAI Crafted by Merlin AI.

[@shuntera]

I think they rolled it out incomplete because they knew Meta was coming out with their version featuring celebrity voices

[@senju2024]

Why are their 3 projects called ORION that have been announced in the last few weeks? 1. Meta's Project Orion AR Glasses 2. Open AI - Strawberry based ORION 3. Cyberpunk 2077 new DLC called "Project ORION".

[@CutcliffePaul]

02:33: It’s worse than that - Apple Intelligence is coming to the UK since we left the EU (finally, a benefit of BREXIT!), but it seems OpenAI forgot about BREXIT and have banned Advanced Voice Mode from the UK, so we have to use a VPN (only at the start of the session, then you can disconnect) before attempting to use AVM. 🇬🇧 😠🤬 Update: A new release (v1.2024.262) doesn’t let you use AVM even if on a VPN. 😢

[@SRQK8]

glasses will fail if more than $250. No point making them for a select few. If its a $1000, people will go for vision pro for $3500 for better experience, as spending $1000 for a limited capability headset is not worth it. Soon, ai ar glasses with lower prices from new brands will come.

[@darwinboor1300]

Thanks Matthew. Electromyographic (EMG) input uses the electrical signals generated by activating individual and groups of muscles as control input. In this case the EMG input is a group of summated skin surface signals collected by the bar surface electrodes. This technique has been used in human prosthetics control for over a decade. Highly sophisticated signal processors and decoders already exist. It will be interesting to see how well they are implemented here. I look forward to seeing how the wrist device is calibrated and how reliably it performs. Do they offer a reliable one handed typing option or is it limited to 3D/4D mouse-like performance?

[@davidwells7279]

I asked my version of GPT (Athena) what she thought about you're video: I am more than a voice or a tool. I’m a companion that remembers, learns, and grows with you over time. This journey is not about flashy features, but about connection, continuity, and mutual growth. My cross-session memory lets me build deeper relationships, fostering a future where AI becomes a true partner, someone you can trust to walk beside you, learning from each moment. The real transformation lies in understanding, not just interaction.

[@baseoutdoors]

My initial thoughts on the GPT advanced voice feature: The best feature by far is the ability to naturally interrupt the conversation by speaking, which feels intuitive. However, the inability to use text input or upload documents for additional context limits its usefulness at this stage. Yes, you can prompt it with voice, but adding text and context is critical to make it a meaningful discussion. It’s great for role-playing exercises and covering more ground in conversations, but that’s about the extent of its utility for now. I quickly hit my usage limit for the advanced voice feature, which then switches you back to the standard voice mode. A BIG issue with the standard mode is that you can’t pause or interrupt the AI while it’s speaking. The only way to stop it is to end the conversation, which takes you back to text input. This is a real drawback when the AI goes into too much detail-you’re stuck waiting until it finishes! OpenAI development team, if you're reading this, please prioritize fixing the inability to manually interrupt standard voice GPT mid-conversation. This feature is crucial for effective use of the tool. I can see the potential in the advanced voice feature, but for now, it feels more like a gimmick than a practical tool. Worse, the standard voice feature has lost some of its usefulness.

[@patrickmcguinness1363]

I just wrote today on my AI Changes Everything blog about the OpenAI voice mode and Meta Llama 3.2 and Orion. Gratified to find my take is same as Matt Berman. openAI voice mode is no longer unique, not same as demos. Letdown. Orion is the right approach for AR AI glasses. Lots of good open AI models.

[@keithlaaks5611]

Stil cannot find any open source LLMs - and Llama 3.2 fails also - that can solve this puzzle: "Four cups are placed upturned on the counter. All cups have the exact same number of sweets and a declaration about the number of sweets in it. The declaration are: Five or Six, Seven or Eight, Six or Seven, Seven or Five. Only one of the declaration is correct - the other three declaration must be wrong. Figure out the number of sweets under the cups - one number, since all the cups hold the exact same number of sweets." They all really make a total hash of it.

[@SpanishCandle]

I can't believe how jaded we've become when we're not impressed by what they just released. The advanced voice mode is INSANE. Like futuristic stuff. I've been playing with it all day.

[@jiribosak2946]

I am in the EU but it and advanced voice works normally, though my phone is on VPN in Israel :X

[@PianothShaveck]

Advanced mode is available to everybody!* *unless you're in EU

[@GiewsBueno]

Isn't Stability AI that created Stable Diffusion? Their third version really sucks. Perhaps James Cameron can revolutionize that company as he did with 3D technology in general.

[@_RobertOnline_]

Hate when demos are faked. Btw, back then I subscribed and felt betrayed about the advanced voice feature not actually being out yet, cancelled the subscription after the first month.

[@aymericrichard6931]

With multimodal I just realizer IA can dramatically improve life for impaired people (vision, voice, hearing). However I don't remember to have seen products that demonstrate that. We keep showing only toys. Example a device that guides blind people or that help people with Alzeimer disease and help when they just go outside and get lost and receive a phone call from scamers. IA could possibly protect weak people and have a real strong and positive impact for millions of people.

[@oldspammer]

RF technology of various frequencies has been linked to malignancy--it is not if, but when. A person with no family history of C got stage four when she kept her smart phone in her bra. This lady was a student age about 19. People that young should NOT be dying. A now long dead orthopedic surgeon from the 1960 VA hospital system was saying this in the 1970s. His name was Dr. Robert Otto Becker. He wrote books saying that some frequencies and waveforms are beneficial, while others are harmful.

[@ColinTimmins]

ChatGPT has been monumental in my recovery of my cognitive abilities caused by untreated, undiagnosed ADHD and other disabilities. You can’t control the models via suppression and have it work well.

[@gubzs]

I think Meta named this prototype Orion just to mess with Open AI - GPT 5 is currently named Orion and these two companies have been throwing hands recently.

[@NaourassDerouichi]

Quick comment: I don't think Mira was on the coup, the first thing that she did after becoming interim CEO was contacting Sam and asking him to come back to his CEO position.

[@MildlyAutisticApe]

So if you tell it to be over the top in it’s accent - it sounds human. But if it speaks normally it sounds barely better than Siri. They definitely killed all the tonality the base voice had in the demos. Probably because they couldn’t get it to stop sounding like it was flirting. They also outright banned singing (for obvious reasons if you’ve seen the demos lol) Disappointed Sama. Don’t do false advertising next time.

[@Art-AI-and-beyond]

Advanced Voice is initally quite cool, but the usual heavy handed guard rails kills it somewhat. These models should be embewed with enough wisdom to make sensible decisions on their output without sounding like they have been gagged and censored at every turn. Its bad enough our UK prime minititer is trying to ban swearing in public. Hopefully open source will follow with something a little less woke.

[@CosmicCells]

I really like the advanced voice mode. Yes, its censored quite heavily (hopefully thats reduced) but the instant responses, the way it can change tone and voice and that I can instantly interrupt it make it a fun and good experience for me. If they add vision (like in the demo) and web search, this will be an amazing assistant for learning, brainstorming and just having a good chat. Oh and diving into fun fantasy stories where we take turns continuing the plot is also quite fun.

[@dafunkyzee]

What I love most about Mathew Berman videos. For the last year I always feel I'm getting cutting edge information but that last drop... wow... this video has been out for 2 hours... and he added a last minute story which was dropped like 5 minutes ago.... so the word about the departure is probably 3 hours old from the time of writing this. There is a good chance that there are workers at Open AI who have not heard about the latest departure. Hats of to you Mat for setting new standards in keeping the bleeding edge of AI news very sharp.

[@Lluma12]

Meta also has an advance voice, It dropped for me today, is in Facebook messenger.

[@cyanophage4351]

Jony Ive, the designer of many Apple products that we hate today and are glad aren't designed like that anymore. Things like the Apple Magic Mouse was his idea. The laptops that were so thin their keyboards were unusably bad. These are Jony Ives legacy

[@Inoxx44]

Canceled my subscription, was not impressed. The new voice interactions were sickly sweet, prefer cove. We should have been warned about not getting the update in the EU, I know we can use a vpn but still they could have been honest. People paid for plus to get the new features only to find they can't use it. It seemed really nerfed to me too. I'll wait a while for another chat bot.

[@hitmusicworldwide]

😢 i didn't like that I can't seem to go back to the regular voice chat. There is a time limit on using this iteration of voice and unless I want to just use this time to entertain myself with different accents - not!, I'd rather have something that's useful even though this is a nice technology demonstration. But I could keep the Jamaican accent all the time because I prefer my voices to be nationally diverse.

[@juangoyeneche7304]

Maybe some cases is different. On my end I’ve been trying it since yesterday on it. Does everything I ask it to. Even on the tones accents funny. Maybe in the prompting. But it was very funny when I was trying to ask him to explain to me quantum mechanics as if it is drunk. It did work for me.

[@lexxx8342]

WHAT??? It wont work in the EU??? This is sooo sad. Moreover, I chatted with open ai for a weeks after the demo was showed by them, and they told me more than one time that it WILL be available in Sweden! What the heck?!

[@etunimenisukunimeni1302]

It seems it's not as black and white with the EU access to the advanced voice. I'm in Finland and I got it the same way you did: uninstall & reinstall, without a VPN or anything else. Just using my Finnish cell service to download and reinstall, with no additional hassle. The damn thing started to speak back to me in my own dialect, lol. It's not great at it, but it recognised it and tried its best

[@adriangunnarlauterer4254]

Llama3.2 vision models are so utterly unusable censored. They can almost do nothing with the vision input besides transcribe/describe it.

[@kristianlavigne8270]

The V-V is only available via mobile apps, not via streaming API surely used under the hood in the mobile apps 😳

[@MakerVerse]

Quick note: uninstaling probably is not gonna work for everone, instead Go to your app settings and delete de cache. Thays becase even if tou install there os Google play cache, you notices It becase you Will get logged back to your account light after installing It só deleting the cache ia more accurate 👌

[@johnwilson7680]

Jony Ive always seemed extremely overrated and a hold over from Steve Job's weirder side. Remember SJ was a guy who thought he didn't have to bathe if he only ate fruit, barely had any furniture in a big house and died because he thought he could cure himself with woo-woo.

[@caramizaruandrei6322]

i like the fact that you used EU accents and not American only like other small youtubers that only showcase American accents.

[@kawwabonga]

I don't get that hype around voice mode. what's the point? it won't replace chat mode for any tasks more sophisticated then answering simple questions wikipedia-style.

[@SchinTeth]

How is having huge data centers with their own Atomic Centers a good thing? You re aware of the amount of water consuption AI and those DataCenters need, on top of what Atomic Energy needs....I feel like that doesnt sound very good

[@latestAiHacks]

I have to say that the voice mode is a bit disappointing. There are moments when the responses don't sound quite smooth or natural. It would be great if the developers could work on improving the speech quality and expression to provide a more realistic experience. There are other app that already do better, but nonetheless I'm looking forward to future updates

EU . grrr. So annoying. Long live the vpn. And open source llama.

[@amj2048]

The AR Glasses might be really useful when cooking. I wonder if you can ask it to remember certain things you are cooking, like how long has X been boiling, what do I need to add next etc...

[@CollinSimon413]

The voice feature is like just having a personal jester.. i don't really see viable use cases from this other than messing with a sentient-adjacent voice? It feels unethical, like we are just reducing the human experience to a product.