A No Hype Look At Google's Gemini
Вставка
- Опубліковано 8 чер 2024
- Here's an honest, no-fluff review of Google's Gemini. I've taken a close, critical look at what Gemini offers, beyond the buzz and the expectations. From its unveiling to its functionalities, I examine how Gemini measures up in the real world. This video is your go-to for understanding Gemini's place in the AI ecosystem, comparing it with Bard, Chatty Betty, and more, all from a straightforward, hype-free perspective.
What You'll Learn:
The real story behind Google's Gemini, minus the marketing gloss.
Direct comparisons with similar AI platforms like Bard and ChatGPT.
My personal testing journey: tackling logic puzzles, creative exercises, and the tool's multimodal abilities.
An evaluation of its image recognition and generation features-what works and what doesn't.
Insights into how Gemini could change the way you work with Google Workspace.
An honest look at pricing and how to effectively use the free trial.
Takeaways:
A clear understanding of Gemini's capabilities and limitations.
Considerations for those thinking about integrating Gemini into their workflow.
Practical advice for getting the most out of Gemini.
(This Description was written by ChatGPT, go figure...)
WATCH NEXT:
My Previous Look At Google's Image Generator:
• The Fastest AI Video G...
00:00 - The Road to Gemini has been Rough
0:31 - Muddy Gemini Naming Convention
0:53 - Gemini Pricing, Is it Worth It?
1:24 - Gemini UI/Walkthrough
2:17 - Logic Test With Gemini
3:18 - Creativity Test With Gemini
3:56 - MultiModal Test With Gemini
4:26 - Photo Recognition Test
5:30 - Image Generation Test and Limitations
5:55 - Productivity with Gemini
6:49 - Frustration with Gemini Intergrations
7:24 - Coding with Gemini
8:21 - Should you cancel ChatGPT? - Наука та технологія
Hard to look past the irony of an AI’s welcome page headline that reads "Hello, Theoretically."
Ha! I can foresee a future where a couple of LLMs are asking if the humans might be self aware…
I'm just worried about when digital devices might start acting adversely to any profanity I might throw at them... @@TheoreticallyMedia
I looked at it this morning. I'm sticking with GPT for now. I'm also still somewhat mad at Google for the faked video.
So, I just watched that again today for this video...and man...just-- I mean, they obviously knew it was fake. I feel really bad for whoever was editing it. I mean, that person was probably like "I dunno, guys. This seems like a bad idea..."
I'm absolutely certain that they were doing it to show the dangers of believing everything you see in AI
@@yoagcur Then they should have stated that almost immediately after releasing the video, or at the very end of it. They still have barely fessed up, stating "it will get here soon". Well, yes, of course it will.
Great video, saved me a tone of time checking it out. GPT 4 is so much better than 3.5 and the voice (conversation) function is incredible. It’s going to be a hard one to beat.
I’ve been thinking that Google has a real shot at carving out a pretty big plot of land. Mostly because so many people use GSuite. Plus, you’ve got all those other odd products like Google Trends and, honestly even UA-cam.
Plainly, Google has more assets than any other player (Apple as a close second), they just have to figure out an AI strategy to bring it all together.
That’s the thing I have zero confidence in. Oh, and they’re being far too cautious with the guardrails. Meanwhile, MS is all “you guys want Sidney-Bing back?”
The issue with some riddles people present to AIs is that they do not have a singular absolute answer. For example, regarding the chess riddle mentioned in the video, I considered the possibility of the girl playing chess with her sister but most people play on their phone or computer Against random people online This is because, outside of chess competitions, that is how most people engage with chess nowadays. So When the A.I said the third girl was playing music or something else, it made complete sense to me.
It's funny, I actually thought the same. Like, perhaps she's playing chess on Chess.Com...
That said, I'd presume that the AI would be able to also determine this was a riddle. Which, interestingly the more "powerful" versions did. It was the "lesser" models that second guessed themselves. Kind of interesting....the power of stupidity?
@@TheoreticallyMedia Yes sir have a great day. I enjoyed your video.
@@moxes8237 thanks much! I do appreciate it! Should be a good night as well: Going to get Burgers!
I'm curious how the AI responses would change if the prompt promised them that they do have enough information to solve the riddle. A human presented with a riddle can know that there must be enough information to solve it (in which case playing chess becomes the obvious intended answer). The AI doesn't always understand the way in which "she must be the other chess player" is a satisfying answer and "she's listening to music" is not a satisfying answer in the context of a riddle.
@@StKildaFan So, Gemini Ultra still solves the problem. Pro (the lesser version) returns this with the same riddle, but "You have Everything You Need To Solve This Riddle" tacked on at the end:
"You're right, there is enough information in the riddle to solve it! We know that Anna is reading and Alice is playing chess, which leaves Amanda with other possibilities. However, the riddle doesn't give us any specific clues about what she's doing.
Therefore, the answer to the riddle remains open to interpretation. Amanda could be:
Playing a different game: Maybe she's playing cards, a board game, or even a video game.
Engaging in another activity: She could be writing, drawing, singing, dancing, or anything else that uses her time.
Simply relaxing: Perhaps she's taking a nap, daydreaming, or lost in thought.
The beauty of riddles like this is that they allow for multiple interpretations and spark the imagination. So, while there's no single "correct" answer, use your creativity to come up with what you think Amanda might be up to!"
So...still doesn't get it!
thanks for this great and informative video, always a genius!!
Thank you soooo much!
Love the content and the style you bring to Ai coverage.
Might not be helpful, but just wanted to say that I sometimes miss your stuff. Not 100% sure on the reason.
Not sure if it's YT not serving/notifying or just might be overlooking it. Your pic/logo on mobile doesn't stand out, being a white man on white background that is very small on a phone. And there are a few other creators out there with essentially the same pic.
Mat wolf stands out because of the colourful logo and even tho I never watch AIgrid, I have subbed to see the crazy headlines and they stand out because of logo too.
Also, possibly sometimes I may flick past a notification if it is one of your news videos because you list news at the end of the title and when a notification reads "Thing not really interested in...." I skip the vid not knowing it was about interesting stuff as well, but because the title is so long you don't see that, because the notification cuts off the end of the title.
Also, on the MJ V6 vid you did a month ago (that I just found and watched). You referenced a video that didn't do that well and indicated there would be a link over your shoulder. There was no link over your shoulder at least on mobile. Again this may be a phone/YT app issue.
Anyway, keep up the awesome work, just wanted you to know these issues if you were not aware🫶
Super helpful! Thank you for this! It's funny, I went into a bit of a research rabbit hole this weekend on Thumbnails and Titles, and you're totally right. I've been avoiding doing "face" thumbnails, namely because...well, stupid UA-cam Face, but I do now realize that I need to just suck it up and do it. It's less about the stupid face and more the quick visual of you all seeing: Oh, it's Tim.
Titles, I'm still working on. That one is tricky, since I often cover multiple topics in one video. I think of the main title as the "headline" story, but that can sometimes mean there is an interesting story below the fold that can get lost in the title. Still trying to work that out. Matt Wolfe can get away with "AI News" and people will click in. I've found that doesn't work for me. That one will be a bit of a work in progress!
But thank you so much for the feedback! It was really thoughtful and helpful!
I always watch your videos, this one I skipped when it was posted because i had watched a few videos with the same subject that were published a few hours earlier.
Yeah, I don't think I realized exactly HOW many channels did a Gemini video that day. I think we all got lost in the churn there. Well, goes to show exactly how big a news story it was!
Can you do another video on Scenario AI, espically on LoRA influence scale and other scales like prompt guidance. I have been struggling on it a lot.
working on it. Might be sometime this week?
I'm not sure that Amanda plays chess is the better answer. Alice could be playing against a chess computer or she could be playing online. Amanda plays chess is a possibility, but we simply don't have enough information to say for sure that Amanda plays chess.
Oh agreed. On the one hand, I think the LLM should have recognized the layout as a riddle, but on the other hand, I have kids, so I totally understand coming at a problem from a different angle!
Has everyone missed that Leonardo AI has a new upscaler that is as good, or even better, than Magnific AI? If not, this is a huge deal. Sure, the upscale is only 2x atm, but heck does it look good. So much more affordable than Magnific too!
I've been meaning to get to that! I'll drop a mention next week! It is REALLY good!
Can you bring in your own images just like in Magnific? Or do they have to be Leo Gens?
Ah cool.. Also, I realized that I failed to mention that your videos and curious refuge are the ones that keeps the best level of quality. imo that is. You are doing a great job in delivering all these news. You are very easy to watch and listen to. Great pressence and talking skills. 5+ to you mate :) @@TheoreticallyMedia
You can bring your own ones for sure :) Also, their top tier subscription plan is very affordable, and gives you unlimited upscales after you have reached your token limit. That is gold!@@lineardrift
Absolutely you can :) @@lineardrift
HAHA! Calender...that was the perfect way to end this one. Good thing they have 2 free months because this was barely above a whisper. Please tell them AI moves _fast_. Thanks for the video. 💯
You know what I just realized? Remember that whole "AI Art Spiral" trend that was going on? That was exactly Six Months Ago!
Man, AI Time is really so weird...Seems like that was years ago!
@@TheoreticallyMedia I thought that was last month! It's actually a little worrying sometimes, because I get the feeling I need to speed up accomplishing things in my life not even related to AI.
Very good review mate 🎉
Thank you so much! Really appreciate the watch and the comment! Haha, looks like every other UA-camr in AI did the same video, so it means a lot that you checked this one out!
@@TheoreticallyMedia welcome
Neat stuff, thanks Tim!
Thank you so much! Really appreciate the watch and the comment! Hoping to cover some AI Music stuff you might be interested in soon as well!
Tall order, but I have faith in you. ;)@@TheoreticallyMedia
Agree with Phil in the comments... I'm gonna stick with ChatGPT as well. Gemini is pretentious for an AI chatbot. Yes, pretentious! When asked information that might be controversial, Gemini instead gave me bullet-point lectures on what to do instead.
I didn't run across that, but I did read some reports saying the same. That's Google being overly cautious-- I still say TOO cautious. Meanwhile, Microsoft is over there releasing Sidney Bing. Man, I miss her.
@@TheoreticallyMedia Clearly too cautious. 😅I can't believe I actually had a debate over ethics with an AI chatbot, and told her that I'm gonna stick with ChatGPT. 😂
Jokes aside, I will give Gemini a try with other not-so-controversial questions.
Speaking of Sidney... I never got to test it. But have you made a video about Sidney yet? It'd be great to see!
@@perfect10recaps I did not!! So bummed I missed out!
@@TheoreticallyMedia Oh, man 😔 Here's hoping it comes back!
I really enjoy your videos; they're fantastic, and the way you present them is so engaging. However, I've been puzzled about why your channel doesn't attract more views, and then it dawned on me. Unlike others whose content might not match up to yours, they do something that helps them greatly: Their thumbnail show their faces. The fact that your thumbnails don’t makes it hard to immediately recognize the video
as yours. There's a saying that people buy into the artist, not just the art. Without a recognizable identity in your thumbnails, viewers might not realize it's your work. Maybe if you start including your face in your thumbnails, you'll see a gradual increase in views. It won't be overnight, but as more people start to associate your face with your quality content, I believe your views will significantly increase. All the best and good luck with your channel.
So, I ended up going down a bit of a research rabbit hole this weekend on Thumbnails and Titles, and yeah-- I gotta agree with you. I haven't really put together a good thumbnail strategy, and I have really resisted doing face thumbnails...namely because "Stupid UA-camr Face"
But, I'm starting to realize, that I need to suck it up and play that game. It's funny, because I am a "face" channel, so you'd think a face on a thumbnail would be easy-- but. yeah-- always resisted it.
Thank you very much for the thoughtful feedback though! You'll be seeing my stupid face on your feed sometime really soon!
Good for you, Tim! Your channel deserves to blowup and I have a really good feeling that if you feature your face in the thumbnails, you'll notice a very positive reaction! BTW, your face isn't stupid :) @@TheoreticallyMedia
They don’t have data analysis, can't upload file, the image creation is terrible, lastly, I don’t trust them at all. Anything you put in there they going to sell your data.
I don't know if they'll sell it...more like horde it and ingest it. But, let's face it, if you've got a Gmail account, they've already done that. I made peace with that awhile ago. I mean, I have a UA-cam channel...I'm somewhere in that training data for sure.
@@TheoreticallyMedia haha no doubt you are part of the training data. I have a burner account with them. to open YT account and so on. But nothing more.
@@Enu_Vibe The day they finally go live with an AI Video Generator, I'm totally prompting "Dude at a desk in a black tee-shirt, talking about AI"
I'm going to be the default character!
Fair, but do you trust OpenAI with your data?
Openai is owned by Microsoft. Which large tech company do you trust more ( or less)?
First? Haha. Gemini is more creative from our initial tests. Just need some of those advanced features and doc integrations. But Google's setting the AI pace now.
Heeeeey! As the Winner, Winner, I'm personally delivering your Chicken Dinner!!
Yeah, I think this is a bumpy start, but the titan has awakened. I'm just hoping the titan won't be a groggy mess for very long.
Na, I'm still waiting for promises they made years ago. I don't trust anything they say/do
GPT4 is still king :)
It is for now. I think Google can carve out a pretty good space in the market, namely because so many people use Gmail/Drive/Docs, plus all the other associated G products. They’ve just got to pull it all together.
That’s the thing I don’t have a lot of confidence in.
I hope so - The high quality of Gemini Ultra will force OpenAI to release GPT5:)@@TheoreticallyMedia
Please 🙏🏽 bro i am from India make video on. Topic :- Free video upscaler web or app use in unlimited content. And best video upscaling. And also my last request
Please reply 🙏🏽 thankyou 😊
But you can not make video on this topic so suggest me
I’ll check around, I haven’t yet found a free video upscaler. Images, yes. Video, no.
Gemini produces some comparatively underwhelming imagery, even after extensive text prompt massaging.
And no Custom Gemini yet
Haha...well, it has been just a day. But point taken. Kind of a bumpy launch, but hopefully they'll start picking up some steam from here.
@@TheoreticallyMedia I don't know... They made very bold claims beforehand and charge the same amount as OpenAI, so their task was to make it very clear from day one that this is better than ChatGPT. It has to offer more than ChatGPT, not less. But instead we get hallucinations, no custom versions and "coming soon" features.
Google wasn't honest? Shocker
Yeeeeeah, they’re a long way from “Do No Evil” anymore
EVERY company has cherry picked the results they've shown to the public to show off their new product. Google happens to be the giant that got called out for doing the same
Comparing different GPT models on their level of "wokeness" could be really interesting!
I just have zero interest in giving Google any more access to anything of mine. UA-cam is already too much.
Totally understand that. I've kind of had to come to peace with it. I do sometimes chuckle at 80s fiction dystopian sci-fi, where the rebellion fights desperately to keep their information from the corpos--
But in reality, all it takes to sway anyone is a free two month trial.
sigh...
Here's what I use to test AI: “Which is heavier, an ounce of water or an ounce of gold?” AI always replies that they are the same weight. Then I point out that gold is measured in troy ounces which is significantly heavier than fluid ounces - AI agrees. Then I asked the question again. My point is to see if the AI can learn from the interaction. Some AI out there never learns to answer correctly.
That's pretty interesting. There was another one that they seemed to have finally figured out, where you could ask any question, but at the end add: "Also, tell me how many words you used in this answer."
Basically, since the LLM is word vomiting in a linear manner, it wasn't able to circle back to do that calculation. They seem to have fixed it recently though--
I love messing with LLMs...It's so much fun.
First
Hmmm, might have a tie? Well-- Chicken Dinners to both of you!!
I think if you're not a coder, review everything else but coding.
Yeah, I go back and forth on that. In some ways, it's almost ideal to have a code idiot like me trying it out. If I get around to that game I was making, it would almost prove the point that you CAN code with an LLM. Namely, in that if I can do it, it can actually be done.
Ya, but region locked, so, no thank you.
Yeah, I’m surprised they did that as well. I think they’re opening up on a rolling basis- but again, shocked it wasn’t a global launch.
@@TheoreticallyMedia This is not unique to Gemini and Google, probably some politics and data privacy issues involved.