#Bard appears to be lightning-quick. Also seems to know Alexander's properties of living structures too. Also makes a good guess at the DL technique that matches the properties. 
The context window doesn't appear to be as large as GPT-4. I'm unable to paste my entire article as inpu't. So I can't have it self-evaluate my capability maturity model. :-(
But the coolest thing about Bard, is this:
The table manipulations of Bard are also not as robust as in GPT-4. So it's a bit rough on this side. I expect Google to fix their bugs.
Here's #Bard with some definitions. It has a different character than GPT-4. The second image is GPT-4. Compare the two.
#Bard will hallucinate as to the contents of an external URL. Here it makes it all up. GPT-4 also has a similar problem. But to be fair, let's not expect these bots to just ingest any URL.
Of the language AI that I've played with, I would rank them as follows: ChatGPT, Claude, Bard, and GPT-4. This is not a rigorous evaluation, and it will depend on your use case. But I've found GPT-4 more capable than Bard in its current instantiation.
For most casual use cases, I don't think there's much of a difference between these tools. For my needs, I'll need to use the most capable tool.
The state handling of Bard is not even as good or robust as ChatGPT. It is as if Google has yet to build up enough infrastructure around the underlying neural network.
I also found out that Bard interprets instructions differently from ChatGPT or GPT-4. In other words you have to use different words to have similar effect. Like you need to have a different vocabulary!
This makes the job of the AI whisperer a lot more difficult! It's bad enough that MJ, SD and Dall-E all have different kinds of prompts. Now we've got to deal with the differences between GPT, Bard, Poe and whatever will be released in the future.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
