10 Shootouts: OpenAI vs Google / Who Won? You Might Be Surprised!

This week, the AI wars reached new heights with major events from both OpenAI and Google. Who came out on top? You might be surprised by the answer as we break down the key highlights and compare the performances of both tech giants.
In this video, we cover:
The OpenAI Spring Event and the new 4.0 model.
Google's sprawling I/O event and its key announcements.
A head-to-head comparison of presentation styles and content.
The battle of AI capabilities: OpenAI’s multimodal model vs. Google’s Project Astra.
Google’s Gemini Pro with a 2 million token limit.
Integration features: OpenAI’s new desktop app vs. Google Workspace integration.
Google’s new AI video generator, "View," and OpenAI’s comparable tool.
Google Music Lab vs. OpenAI’s future AI music capabilities.
The hardware showdown with Google’s TPU Trillium.
The debut of Google’s "Gems" and "AI Teammate" vs. OpenAI’s GPTs.
Join me as we analyze which company took the lead in each round and who ultimately won the AI showdown.
FlexiSpot Amazing Brand Day Sale! FlexiSpot provides all kinds of standing desks to meet your demands. If you prefer a Rock-Solid Stability standing desk, then the E7 Plus is your top choice! Use my code “24BDYTB50” for extra $50 off on E7 PLUS! Don't miss out on designing your dream set up with FlexiSpot.
FlexiSpot E7 Plus standing desk:
bit.ly/4dIf2oj (USA)
bit.ly/4dFNDmF (Canada)
Let me know your thoughts on which event impressed you the most in the comments below!
CHAPTERS
0:00 - Intro
0:36 - Round 1
2:21 - Round 2
3:41- Round 3
4:35 - Round 4
6:33 Studio A update & Flexispot Sponsor Segment
9:13 - Round 5
10:17 - Round 6
10:47 - Round 7
11:39 - Round 8
12:34- Round 9
14:21 - Round 10

[@petratilling2521]

I fell asleep 4 times trying to watch the Google thing.

[@TheoreticallyMedia]

Seriously. It was a struggle the second time. I could see how they were organizing the flow of information, but even then it still felt so scattershot. You want AI? Here’s ALL THE AI!!!

[@wtflolomg]

Point of fact: Udio uses GPT-4o (previously GPT-4T) for it's lyrics. It should count for something.

[@TheoreticallyMedia]

Does it really? I did not know that...well, extra point to OAI!!

[@wtflolomg]

@@TheoreticallyMedia That bit leaked out on the Udio discord site update announcements, where they told us their generations should be a little bit quicker in creating lyrics because they switched over. I'd really like to know their underlying model for the audio generation, though...

[@narottamzakheim5051]

@@wtflolomg Interesting becuse Udio is actually really good at writing lyrics. With GPT 4o I expect it to be even better.

[@Alderberanirgal]

12:39 Gems vs. GPTs...

[@TheoreticallyMedia]

Gems is still a stupid name!

[@rickwelin9773]

THANKS FOR STRAIGHTENING THE PAINTING TIM. Major OCD trigger for the first half LOL 😂

[@TheoreticallyMedia]

Haha- sorry about that!!

[@dylan_curious]

Thanks for these updates! I downloaded the Mac app for desktop, but it’s still really limited. I couldn’t find a way to screen share in real time or anything yet, but it was nice for doing searches with a shortcut.

[@TheoreticallyMedia]

Yeah, I'm kind of just leaving the chatbar up and treating it like a floating window. A bit easier than clicking over tabs. It's funny, between the new desktop app, moving over to Arc browser (have to tried that yet?) and all my searches now on Perplexity-- it does feel like a lot of my workflow has changed for the better! Maybe not SHOCKING (ha!), but I'll say there is a feeling of less friction.

[@FilmFactry]

We won!

[@TheoreticallyMedia]

We ALL won!!!

[@circelink]

As always, you give us the goods Tim! Thanks!

[@Glowbox3D]

All amazing, and we all benefit from this insane tech blitz. I honestly think the most base interesting thing to me is the low latency realistic voices of Omni. I don't love the valley girl flirting, and I would absolutely prompt that out of my model when using it, but it's really 'the magic' for me right now. Keeping a bleeding edge AI's ears open while I work, and being able to just talk and bounce ideas off it real time no latency is just science fiction become reality stuff. I use GPT-4 now in that manner (sometimes Gemini if it's software related), but the latency and mono-drone voice is far less magical.

[@kgnet8831]

The full Google IO playlist (kzbin.info/aero/PLOU2XLYxmsIJQPs-2nnII-n59vYoxKsmq) has 109 videos and is long (I did not bother to calculate), even the concentrated list with only the 12 keynotes (kzbin.info/aero/PLOU2XLYxmsIJlQyJct5iZj_t7yM9Gjqzj) is about 9-10 hours long... 😺 That being said, there were so many new things that were important (like Gemini Nano in browser) or all these new open-source models etc, and the importance of a local device model, that by sheer volume google crushed OpenAI. The most important thing for me is that OpenAI is like Apple. It might be shiny, but living in a country with strong data protection laws, most of OpenAIs stuff is not usable at work, and the rest is uninteresting if you have an android phone and linux on your desktop... Then there is only 4o left and the promise of free for all, what is practically wrong, because you only get a few requests and after two follow ups you get blocked, so also not very usable...

[@noobicorn_gamer]

OpenAI seems to be more very startupy presentation vibes that does not make it all pretty or whatever to just show how good their products are (and are becoming) while Google seems like they are trying to hide their lack of progress behind flashy performance presentation (Marc… really?) OpenAI is clearly targeting the general public and Google is just… all over the place (actually just trying to cater to their shareholders I guess). At least for public perception of things, OpenAI is sure doing a great job. All they have to do now is just announce they are building a AI phone and things will get very wild.

[@TheoreticallyMedia]

Haha, I love Marc, but yeah-- agreed. I actually think Google would have done better with smaller events through the year, as opposed to the ONE BIG SHOW. They're a big company with lots of different limbs, and when you're trying to show off everything at once-- it just comes out a big jumbled mess. I actually think of Disney-- they have their big D23 event, but they generally will do a "Star Wars" day, and a "Marvel" Day. Basically, they parse out their big announcements, otherwise it all turns into a wall of noise. ...I mean, I'm excluding the fact that Disney has turned Marvel and Star Wars into a wall of noise, but I think you get my point!

[@Relivino]

The winner is... Stable Diffusion!

[@jeffg4686]

DALLE is definitely better now than 6 months ago (at following the prompt) - graphics was always good. What we need is the text-to-3d. I've seen what Nvidia has been doing in that space - Wow (probably way better as I type). What's the hold up? I'm assuming they're waiting until they get it good enough for people not to switch platforms. But who knows - there could be like 10 reasons why, I suppose. Too easy to start creating and 3d printing weaponry - for one, but that's pretty easy right now with a little more time. At the rate computers are moving, Trillium's 4.7x increase in speed is NOT IMPRESSIVE. I was expecting 60x increase

[@TheoreticallyMedia]

Agreed! Dall-E is looking really good these days. I just want the ability to "chat" with it via ChatGPT to make inpainting and outpainting. That is something they've strangely neglected. We do now have in/out painting selection boxes in DallE, they just aren't very good yet. But, as a straight up image generator, I think it looks pretty good. WAYYY better than v.1! And yeah: 3d is the next big frontier! I've got my eye on Midjourney to be the first out the gate there.

[@jeffg4686]

@@TheoreticallyMedia - yeah, I've been messing with DALLE again lately. It's okay at inpainting - sometimes you just have to try a few times with different language. It's not great tho... Haven't tried the outpainting and didn't know it existed in DALLE - will have to take a look.

[@Crazy_Truth]

Yup, I sent only 4 to 5 messages, ( GPT 4O ) and then it showed that I had reached the limit. How can anyone call this free!!?

[@Keenan686]

how can free users get access to this thingy ma jiggy?

[@Crazy_Truth]

@@Keenan686 they will notify...;

[@TheoreticallyMedia]

Yeah, I expect super low caps on the free tier for awhile. You'll have to wait for the hype crowd to thin out a bit. Actually give it a shot on the weekend-- you might get better results.

[@bootkins]

Hopefully I misunderstood, sounded a lot like shade was being thrown at Marc Rebeillet

[@TheoreticallyMedia]

oh, no no-- big fan! If anything, I'm STOKED he got to play that show-- BUT, there were for sure a lot of confused audience members. The general online chatter seemed to echo that. There was a lot of WTF'ing. I thought it was great...an odd choice for a show like that-- but a real treat if you love Loopdaddy.

[@freyna]

He wasn't shading MR. His musical skill amazing. His fit, not so much. It was an odd start. They didn't seem to truly know their audience in opening with him. Not his fault. I came there to hear tech news. Focus on that. Why have an opening music act to try and wow people as an ad for their music LLM DJ mode. It was an annoying opening.

[@K.F-R]

OpenAI releasing its desktop app only for the platforms heavily investing in their company seems at odds with Sam's "free for everyone" message. No linux version even though the mac version codebase would be almost identical. Open source is the only safe way forward. The greed/haste to lock people into an ecosystem is palpable.

[@TheoreticallyMedia]

Oh, I feel those walled gardens coming up as well. Been around long enough to remember the free wheeling Napster/Internet Wild West days… But like the cowboys of old…well, I played Red Dead Redemption, so I know how this story ends. …and NO I’m not THAT old! Haha

[@DataJuggler]

0:50 No, I don't own a cell phone, so I can't use Open AI.

[@TheoreticallyMedia]

Ah...smart move. You will be leader of the resistance when Skynet comes for us.

[@twilightfilms9436]

No app in the world will let you sing happy birthday. It’s the song with the strongest IP and copyright protections in history and sync it to a film is quite impossible unless you’re willing to spend several zeros……they’re so strict that it can’t even be used for charity matters……

[@TheoreticallyMedia]

Ahhh- good call! Totally forgot about that! It's been awhile since I've been in a restaurant and heard the "alternate company branded birthday song" that they're legally required to sing, since HB is copywritten!

[@virgilxavier1]

That's actually not correct the happy birthday copyright is not only public domain but it has been ruled to be public domain in court on more than one occasion. Good example of copyright trolling.

[@virgilxavier1]

​​@@TheoreticallyMediaalso the story about a restaurant having to sing a variant shows how misguided the story is because any restaurant would have a performance license which would entitle it to sing any song

[@virgilxavier1]

KZbin will not allow me to post a link but Google happy birthday public domain Court ruling and you will see that in La Court ruled on this in 2015 and the company had to pay out 14 million in damages for fraudulently collecting on a work that they did not own the rights to

[@twilightfilms9436]

@@virgilxavier1 that’s not correct. Call ASCAP and tell them you have to sync in your cue sheets happy birthday…..give it a try, because I’m speaking from what the studios I work for are telling all the time. Never use happy birthday……

[@petratilling2521]

“you likely have access to it” 😢

[@TheoreticallyMedia]

I’m sorry!! I always have to say stuff like that because 5 months from now; someone will yell at me in the comments like; What do you mean I don’t have access?! There’s always this mental game you have to play with wording things in YT videos!

[@narottamzakheim5051]

Trouble with Gemini Pro. I seen Mathew Berman testing it... running it through its standard logic tests... and Gemini pro performed worse than pretty much any other model Ive seen him test. it was unable to process information even on 100k context. It was not impressive at all. Whereas GPT4o blew through his tests and performed at a top level. Google is throwing out big numbers but honestly... I dont see them being effective in using them. Just like you... the feature didnt even work. Google also ties most of their features into their products... which I dont use... many people dont. Useless features for me. Open AI focused on the voice assistant... but GPT4o actually has a huge set of incredible features... they just werent mentioned at all in the presentation. For some reason... they just left it to the website. Basically... Google showed a lot but delivered a little whereas Open AI showed little but delivered a lot. Open Ai has released a much better product that will actually do what is advertised and more. Bottom line... Im gonna subscribe to Chat GPT plus... not Google.

[@trutenantedboderampt]

Me too!

[@CoolZero101]

I also agree with the subscribe to Chat GPT Plus.... but the "I don't use... many people don't" is a bit misleading.... I did a quick search and "As of the first quarter of 2024, Android has a 70.7% market share of the global mobile operating system, while Apple's iOS has 28.5%. Android and iOS together control more than 99% of the mobile operating system market share." I also want google to "tie mpost of their features into their products" as most people will likely do most of their AI interactions with their phone, which is predominantly Google. I hope for the best for both, as it will only benefit everyone in the future.

[@LouisGedo]

👋 You Open AI hater!! 😉 🤪 😜 Since I'm not biased toward or against any AI developer, to me, the better company is the one which offers or is developing the most useful and high quality tools for my needs that I cannot get somewhere else. By this criteria, Open AI hands down crushes and eviscerates Google. I don't need a chat bot that can accept 2 septillion (or whatever) tokens. There's nothing about that type of thing that interests me at all. As for AI music creation, I already have access to superior technology such as Udio & Eleven Labs. If Google wants me to gush, then they need to develop tools I need or want 👍 👍 👍

[@TheoreticallyMedia]

I JUST want them to have an AI assistant that will manage my inbox and calendar! I don't know why that is so hard for them! Actually, having spoken to a few G'ers, the reason is so stupid: The teams don't talk to each other. It's insane.

[@LouisGedo]

@@TheoreticallyMedia 😲 no way!

[@MYFAVORITES5]

For AI generated imagery, I'm still on the Midjourney train (I tend to like their Niji anime creations). If they can figure out turning the images into animation / video, then they would get all my AI money. 😅

[@TheoreticallyMedia]

That's really what I'm waiting for. I don't cover MJ much on the channel, unless they've got something big happening (although I did just do a "MJ for Beginners" video on another channel)-- but, yeah, it's still a daily use for me. I'm counting on them to crack that animation/video that is killer.

[@Nerf_Jeez]

Easy openai w

[@TheoreticallyMedia]

agreed! Google announced a LOT. But at least OAI came out with: "This thing is going to be coming out, but you can also use a bit of it now." Google was just a lot of "It's coming soon..." And-- I'd be ok with that, if not for Google's track record of announcing things that never actually appear.

[@jjay775]

@@TheoreticallyMedia Totally - Low latency stuff you can actually use now beats vaporware every time

[@ECHSBACHS]

: )

[@TheoreticallyMedia]

Appreciate it! This video was a bit of a nightmare to put together! Lots of insane hiccups all the way through, so I really do appreciate it!

[@stealplow8462]

Google was so boring.

[@user-lp7vz9sz7x]

Because you are not engineer 😅😅😅

[@TheoreticallyMedia]

And long. so very long...

[@seventyfive7597]

Google won, easily, since GPT4o was GPT5 relabeled due to it being disappointing in reasoning. 4o was not planned as an incremental update, no one rehauls the tokenizer and adds modality to an incremental model, it's simply too expensive in training. This was supposed to be 5, and it disappointed, and hence rebranded to avoid investor backlash. A loss for a year and a half in the works.