5 ms·
I think they massively screwed up by releasing half baked coding assistance in the first place. I use ChatGPT as part of my normal developer workflow, and I gav
by TX81Z 3y ago
I think they massively screwed up by releasing half baked coding assistance in the first place. I use ChatGPT as part of my normal developer workflow, and I gave Bard and ChatGPT a side-by-side real world use comparison for an afternoon. There is not a single instance where Bard was better.
At this point why would I want to devote another solid afternoon to do an experiment on a product that just didn’t work out the gate? Despite the fact that I’m totally open minded to using the best tool, I have actual work to get done, and no desire to eat one of the world’s richest corporations dog food.
- dist-epoch 3y ago[flagged]
- tough 3y agoWait aren't we all paid to be here?
- TX81Z 3y agoI generally get in that benefit from the time I spent on here to learn about new things that are pertinent to my work. Whether or not I want to keep going back and re-testing a product that failed me on the first use is a completely different issue. Also, it’s a good thing I run my own company. My boss is incredibly supportive of the time I spend learning about new things on hacker news in between client engagement.
- wilg 3y agoWho cares, just check back in a year and see how its going.
- nvy 3y agoYep, the progress will be slow but inexorable on this front. Sooner or later we'll arrive at what I see as the optimum point for "AI", which is when I can put an ATX case in my basement with a few GPUs in it and run my own private open source GPT-6 (or whatever), without needing to get into bed with the lesser of two ShitCos, (edit: and while deriving actual utility from the installation). That's the milestone that will really get my attention.
- nsvd 3y agoYou already can run a local llama instance on a high-end graphics card (6+ GB VRAM).
- tpmx 3y agoAnd it's hilariously bad (in comparison to regular chatgpt).
- Der_Einzige 3y agoAnd slow. They never tell you that quantization of many LLMs slows down your inference, sometimes by orders of magnitude.
- arugulum 3y agoIt depends on the quantization method, but yes some of the most commonly used ones are extremely slow.
- nvy 3y agoYes, I can, but (see my edit) there's very little utility because the quality of output is very low. Frankly anything worse than the ChatGPT-3.5 that runs on the "open"AI free demo isn't much of a tool.
- TX81Z 3y agoPrecisely my point I don’t think a lot of people will go back. Even somebody like me who’s willing to put several hours into trying to see how both work won’t do that for every blog post about an “improvement”. Bard was rushed, and it shows. You only get one chance to make the first impression and they blew it.
- wilg 3y agoI don't think so for stuff like this, it kinda has to be built in public, and iteratively. If it gets good enough they'll surface it more in search and that'll be that.
- TX81Z 3y agoPartially agree with that sentiment but I don’t think it negates my point that they released something inferior because they were caught flat footed.
- LightBug1 3y agoCompetition is competition and I respect that. I'll use whatever is best in the moment. And if chatgpt start trying to network effect me into staying locked with them, I'll drop them like a bad date. Been there, done that. Never again. Ymmv
- jejeyyy77 3y agoWhat? After a year, they'll hear that Bard is really good at code assistance now and then they can try it again.
- telotortium 3y agoBard is fast enough compared to ChatGPT (like at least 10x in my experience) that it's actually worth going to Bard first. I think that's Google's killer advantage here. Now they just need to implement chat history (I'm sure that's already happening, but as an Xoogler, my guess is that it's stuck in privacy review).
- okdood64 3y ago> I think that's Google's killer advantage here. Also it can give you up to date information without giving you the "I'm sorry, but as an AI model, my knowledge is current only up until September 2021, and I don't have real-time access to events or decisions that were made after that date. As of my last update..." response. For coding type questions, I use GPT4, for everything else, easily Bard.
- rrrrrrrrrrrryan 3y agoHave you used Bing? It's great for stuff up until a few days ago (not necessarily today's news), powered by GPT-4, and the results have been consistently much better than Bard for me.
- TX81Z 3y agoIn my testing it was faster with worse answers, and GPT spits out code only slightly slower than I can read it. I don’t care for “fast and wrong” if I can get “adequate and correct” in the next tab over.
- telotortium 3y agoAh, maybe that's a difference - I can read an answer of the size that ChatGPT or Bard in 1-2 seconds
- TX81Z 3y agoI read human language quickly, I’m talking about the rate at which I read code from the internet I’m about to copy and paste. Which is, and I’m my opinion should be, slow. But I agree for normal human language GPT needs to pick up the pace or have an adjustable setting.
- elicash 3y agoI don't think there's much harm. If they ever get to a point where it's reliably better than ChatGPT, they could just call it something else other than "Bard" and erase the negative branding associated with it. (If switched up the branding too many times with negative results, then it'd reflect more poorly on Google's overall brand, but I don't think that's happened so far.)
- redbell 3y ago> they could just call it something else other than "Bard" and erase the negative branding associated with it That’s exactly what Microsoft did for Internet Explorer.. They totally got rid of this name in favor of “Edge”
- bjord 3y agoI assume you're using GPT-4? In my (albeit limited) experience, Bard is way better than GPT-3 at helping me talk through bugs I'm dealing with.
- TX81Z 3y agoYeah, 4
- gwd 3y agoEvery so often I go back to GPT-3.5 for a simpler task I think it might be able to handle (and which I either want faster or cheaper), and am always disappointed. GPT-3.5 is way better than GPT-3, and GPT-4 is way better than GPT-3.5.
- bjord 3y agoYeah, I actually meant GPT-3.5 when I said GPT-3. I haven't personally tried GPT-4 at all. I'm actually happy with Bard, but it seems like I'm the only one.
- gwd 3y agoI mean, I was pretty happy with GPT-3.5 while I was waiting for GPT-4 access. But once you get used to it, it's hard to go back.
- swyx 3y agoi run them all side by side all the time btw https://github.com/smol-ai/menubar/ https://github.com/smol-ai/menubar/