Google Bard vs. ChatGPT - Side By Side Comparison

💻My Stable Diffusion PC Specs: kit.co/AllYour...
📌Stay Up To Date On The Latest AI News: allyourtech.ai
🎨Free Stable Diffusion Discord Bot: / discord
Today I take a look at the all new language model from Google called Bard. How does it stack up to ChatGPT and specifically GPT-4? I test both systems side by side with the same prompts and compare the results. Let me know which system you think comes out on top!
Join the AllYourTech Discord Server and talk about all things AI: / discord

[@AdamGrace]

[updated] I just discovered your channel today but you are brilliant. Im binge watching now lol. Keep up the great work man!

[@allyourtechai]

Thank you so much! I really appreciate the kind words

[@benjaminwaterman9580]

I started using both about a month ago and my results aren't all that dissimilar from your video. I find that I prefer to chat with Bard. The answers tend to be much more "human". I do prefer to use ChatGPT for code, though, especially if I need it to work without much debugging (like in Excel). The context is also helpful in ChatGPT, Bard just flat forgets what I'm talking about. I wrote a couple of novels and pointed Bard at my links and it just made stuff up - but it rotates between two stories. I kept trying and it started to play with me. At the end it said "it's about..." "just kidding, it's about..." "I was kidding again, it's really about" and just kept that up for a while. That was a complete surprise. It not only hallucinated, it also got frustrated with me and got sassy. I think that Bard has a great deal of room to grow and will do so quickly, far outstripping ChatGPT. That being said, ChatGPT is currently shinier than Bard. I prefer Bard for my daily chats, but ChatGPT is simply more ready for most tasks. Having the ability to discuss current events is a strong factor in my preference for using Bard.

[@allyourtechai]

Bard was recently updated to PaLM 2 which significantly increased the capabilities of bard.

[@benjaminwaterman9580]

@@allyourtechai I knew that Google was going to do that, I just didn't know they already had. But yeah, I am excited to see what they do with this tech. I want it on my smart speaker, my searches, everything. I want to converse with this instead of a human when I call the phone company (although a human would probably be more useful). I want it to give me my performance review. Ooo, improve my emails and text messages. It would be nice to be able to chat about my problems and get a useful answer. Humans, man, too many biases without much usefulness.

[@Galaxia53]

Asking Bard to summarize an article about Bard to see if it can access the internet isn't a very good way to test that because it will be hard to tell if it just knew a lot about it self already or if it actually read the article. An obscure website which is unlikely to be included in the training data would probably be best

[@allyourtechai]

Agree, that would have been a. Far better example

[@Galaxia53]

This is just a theory but GPT 4 might be so slow because of a censorship filter. I heard that the AI on character ai used to be a lot faster before it got a censorship filter and the slowness of the typing is very similar.

[@allyourtechai]

That seems like a reasonable theory. The model they are using for “fine tuning” which would include things like censorship would definitely have performance implications .

[@solaawodiya7360]

Amazing comparison. Can you compare Bing vs Bard? Since Bing is powered by GPT-4 and faster

[@allyourtechai]

Will do!

[@NickWindham]

Google Bard wrote Python and Rust code for this afternoon that worked. I think it will improve extremely quickly because Google’s feedback system is better than GPT-4’s and Bing’s

[@allyourtechai]

I agree. OpenAI put significant resources into the coding side of the training for GPT-4. No reason google can’t do the same.

[@ggman69]

Bard isn’t currently supported in Canada!

[@cryptojedii]

When you said "the results that came back from GPT-4 is a complete hallucination..", I darn near spit out my coffee!! Hahaha You weren't LYING!! Talk about a made-up short story on the fly!! 😂😂😂😂

[@allyourtechai]

It’s really funny sometimes when you get back something completely insane lol

[@kamtschatkas]

Bard says the RTX 3090 has 5248 CUDA cores, which is wrong. It has 10496. Chat GPT got that right.

[@allyourtechai]

Good catch, I noticed that after the fact. In my tests, Bard seems more like GPT-3 in terms of the quality of results returned.

[@edwinungerer7989]

they just added plugins so chat GPT is on a whole new level now

[@allyourtechai]

I saw that a few minutes ago

[@tenzintsenpey5274]

you are comparing apple to oranges. We all know that ChatGPT isn't connected to the internet. if you do with bing chat then the result will be significantly different.

[@allyourtechai]

That's why I showcased that point. It's important to note that ChatGPT isn't connected to the internet (that's about to change with the release of plugins)

[@tenzintsenpey5274]

@@allyourtechai fair enough. Do the same when they release the plug in.