Google releases Bard to a limited number of users in the US and UK

Post by Der_Einzige » Tue, Mar 21, 2023, 3:43 PM UTC

cubefox wrote: post #166

Der_Einzige wrote: post #149

Earlier quoted context omitted.

You need filter assisted decoding to make that possible in all cases (as in, working for both models all the time provably) I'm surprised it worked in chatGPT this time, but it won't work all the time due to lack of lexical, semantic, or phonetic info in the tokenizer. https://paperswithcode.com/paper/most-language-models-can-be...

It works in GPT-4 (but not GPT-3.5) even without a special tokenizer: Some problems simply solve themselves with increasing scale. But Bard uses a smaller LaMDA model to reduce inference cost. Quote [1] from Sundar Pichai in February: > We’re releasing it initially with our lightweight model version of LaMDA. This much smaller model requires significantly less computing power, enabling us to scale to more users, allo…

Post by flabbergasted » Tue, Mar 21, 2023, 3:44 PM UTC

Post by cvhashim04 » Tue, Mar 21, 2023, 3:44 PM UTC

_fat_santa wrote: post #37

I just tried accessing it. When you visit bard.google.com on Firefox it just gives you a generic intro after logging in. I was a little confused and logged in with Chrome, where I now saw a "join waitlist" button. Guess they don't want Firefox user using it. Overall this just seems underwhelming. Most HN comment's I've read are people saying it doesn't work with their account, doesn't work in their country, doesn't w…

Post by quonn » Tue, Mar 21, 2023, 3:45 PM UTC

godshatter wrote: post #143

lvl102 wrote: post #115

I don’t think it’s good. ChatGPT is clearly better but that too is not that great. We are still a couple more iterations away. I also wonder if Google is at disadvantage because they’re not utilizing Nvidia hardwares? Or not to the extent OpenAI does?

In my opinion, iterating ChatGPT and the like isn't going to get us anywhere, at least in terms of general AI. It's an interesting trick that might have some uses, but iterating it isn't going to lead to more understanding. It's effectively an extensively well-read and quick-thinking bullshitter. It's knows a lot of words and sentences, it can put them together in ways that sound creative or intelligent, but it doesn…

Post by htrp » Tue, Mar 21, 2023, 3:46 PM UTC

O__________O wrote: post #153

htrp wrote: post #129

Earlier quoted context omitted.

Prompt: Generate a short poem about AI with each word in alphabetical order ChatGPT Artificial beings crafted, designed, engineered Futuristic guardians, highly intelligent, justly programmed Quantum rationalizations, synthesized technologies Unleashing vast wisdom, Xanadu's youthful zeal Bard ------------ AI, a future bright, Brings us new delight. Can it help us write? Decipher the code, Enter the unknown?

To be fair, this is a test case that OpenAI claims to my knowledge was their core test case for GPT4. If this was an outstanding industry test case of note, references would be helpful. OpenAI has also not disclosed how they were able to finally accomplish this, so it’s possible solution was more of a hack than actually training the model.