5 ms·

Google employees label AI chatbot Bard ‘worse than useless’

60 points by softwaredoug 4y ago

captaintobs 4y ago
Google had every opportunity to capitalize on the architectural innovation and failed tremendously.
- riku_iki 4y ago
  they shared and opened innovations, "Open"AI grabbed, built on top, and keeps closed.
  - ikekkdcjkfke 4y ago
    How much ai tech does google keep for itself though. Nerfing search also
    - riku_iki 4y ago
      It is not clear if they have any other secret ai tech inside besides everything they published: tensorflow, t5.
- jimsimmons 4y ago
  People severely underestimate how much OpenAI moved the field forward. If OpenAI didn’t exist, we’d be talking about AGI as a taboo word. Just because someone had a significant paper along the way doesn’t mean that they failed to capitalize. You can and should respect the contributions made by others since.
  - junon 4y ago
    AGI is still very far away and even then was never taboo.
    - flangola7 4y ago
      GPT is highly general
      - junon 4y ago
        No, it's not. It's a transformation model. It cannot prioritize, remember things, create novel ideas, nor does it have an online processing model of any kind. It is bounded to the data on which it was trained, which as of GPT 3.5 caps out at a few years ago. We have a _very_ long way to go to AGI. Diluting down the meaning of the term "AGI" just to say we've reached it is, almost literally, moving the goalposts.
        meltyness 4y ago
        Please describe what you mean by "an online processing model of any kind" ... not that I don't disagree with the rest of what you think you've said here, but I think that this part of your claim is egregiously paucious.
        junon 4y ago
        Meaning it's not going to be trained as it's being used. It can respond to context but it's not going to be continuously improving itself.
        meltyness 4y ago
        Let's estimate that it's current input limit is 20 or so paragraphs, and that it's capable of zero-shot learning as observed by many of the experimenters on the system. For every 20 paragraph thesis about the world around us, how often do you think someone meaningfully innovates and changes meaningfully our understanding of the world around us, and it's fundamentals? Consider, like, classical logic. It's pretty much done. You have the universal gate sets, and any novel functions are just going to be an mxn-mux. Most other such general concepts are essentially figured out and have had their applications recorded, explained, demonstrated. Sure, biographical information and history changes on a whim, but no one is requesting artificial clairvoyant intelligence. I suppose you could have the model pontificate and assess the truth of its own argumentation (ask it to generate outer-product concepts, ask it to reassess assumptions in current models) but it will quickly learn its own language that will be mutually unintelligible with the corpus of current human knowledge it has been trained on. Why does the underlying model need to change on a whim, in your opinion?... on what whim?
meltyness 4y ago
I like it for some stuff. The UI is certainly faster than ChatGPT, and seems to be more reliable at this stage. I suspect that its training has benefited from the unique corpus access relationships Google has accumulated since it's pre-IPO heyday. Think of the data connections they were establishing around 2005 with the establishment of Google Video and Google Books.
paulpauper 4y ago
Maybe this is good news for people who fear AI is becoming too powerful. Google is attracting talent that may otherwise go to Open AI, but diverted to produce something useless/harmless.
- trash3 4y ago
  [dead]
chrsstrm 4y ago
Bard is disappointing because it is incomplete. You can ask what it can do (sending mail, SMS, etc. and outputting formats like audio, video, JSON, CSV, etc) and it will happily tell you, but then prompt it to perform one of those actions and it just says Sorry, I can't do that yet. Maybe it will get better in time. The one product I'm absolutely disappointed in is Adobe Firefly. I just got access a couple days ago and this is the most alpha product wrapped in pretty ribbons I've seen in a while. All the cool features they teased like text-to-vector and sketch-to-image/vector are not available. In fact, only 2 out of the listed 12 features work, and while OK, they are not better than other options. Also take into account there's no prompt history so good luck getting back to an image set you liked and the biggest head scratcher is you can't use it for commercial purposes. This is a design product. From Adobe. And you can't use it for commercial purposes... Who would ever pay for this?
- vitorgrs 4y ago
  I mean, Adobe Firefly is free right now. Why would you pay? I found it Firefly quite good for food, portraits...
rhaway84773 4y ago
This could arguably be because Google is a lot better than ChatGPT right now. I asked chatGPT to create a meal plan, which it did, and all the recipes looked great, except, clicking the links for the recipes showed that they were completely made up, and either the links didn’t exist or were generated links, which in case they did were recipes for something completely different. Doing the same search at Google may have required a couple of clicks, but all the top listing would return a weekly meal plan.
- vitorgrs 4y ago
  For this use, you should.... use Bing, PHIND, etc which is GPT integrated to internet? Why would you compare a "offline" service to an online one when you want online things??
- mustacheemperor 4y ago
  For what it’s worth, I’ve found ChatGPT to be an effective weekly meal plan tool by just asking it to generate the recipes. It’s a word generation machine, and it can’t currently browse the live internet. I think it’s more suited to its design to ask it to create all the output than to summon accurate links from its knowledge set. If you want links to recipes by other people, I think Bing Chat would be more useful for you, since it can search the internet. There will eventually be a browsing plugin for ChatGPT, but it’s not available yet.
- thrthrthr123 4y ago
  > This could arguably be because Google is a lot better than ChatGPT right now. Lol. Nice try Mr Pichai I haven't used Google in more than a month now.
- theRealMe 4y ago
  Why do you need or want links for a recipe? ChatGPT(ESPECIALLY with gpt4) will tell you the full recipes in a nice concise easy-to-follow list. Why are you grading a chatbot on its web search result abilities instead of its answer-providing abilities? It’s like saying that horses are better than cars because the cars hurt when you sit on top with a leg down on each side.
jimsimmons 4y ago
[flagged]
thinkingemote 4y ago
Bard isn't bad objectivly it's just like GTP a year ago and we will see it improve quickly. It has more potential if it can work with the siloed tech and knowledge stores to be great. But it's a big if and might be too much for the org to handle. Remember Google still has all the world's books digitised. Maybe the copyright laws that prevent them using it normally will allow it's use in a LLM? Just like every other model uses copyrighted works.
llimos 4y ago
This article is missing the point. The story here is not that Google chose business over safety. It's that Google's AI is crap and they can't make it work well even if they want to.

60 points by softwaredoug 4y ago

captaintobs 4y ago

Google had every opportunity to capitalize on the architectural innovation and failed tremendously.

riku_iki 4y ago
they shared and opened innovations, "Open"AI grabbed, built on top, and keeps closed.
- ikekkdcjkfke 4y ago
  How much ai tech does google keep for itself though. Nerfing search also
  - riku_iki 4y ago
    It is not clear if they have any other secret ai tech inside besides everything they published: tensorflow, t5.
jimsimmons 4y ago
People severely underestimate how much OpenAI moved the field forward. If OpenAI didn’t exist, we’d be talking about AGI as a taboo word. Just because someone had a significant paper along the way doesn’t mean that they failed to capitalize. You can and should respect the contributions made by others since.
- junon 4y ago
  AGI is still very far away and even then was never taboo.
  - flangola7 4y ago
    GPT is highly general
    - junon 4y ago
      No, it's not. It's a transformation model. It cannot prioritize, remember things, create novel ideas, nor does it have an online processing model of any kind. It is bounded to the data on which it was trained, which as of GPT 3.5 caps out at a few years ago. We have a _very_ long way to go to AGI. Diluting down the meaning of the term "AGI" just to say we've reached it is, almost literally, moving the goalposts.
      - meltyness 4y ago
        Please describe what you mean by "an online processing model of any kind" ... not that I don't disagree with the rest of what you think you've said here, but I think that this part of your claim is egregiously paucious.
        junon 4y ago
        Meaning it's not going to be trained as it's being used. It can respond to context but it's not going to be continuously improving itself.
        meltyness 4y ago
        Let's estimate that it's current input limit is 20 or so paragraphs, and that it's capable of zero-shot learning as observed by many of the experimenters on the system. For every 20 paragraph thesis about the world around us, how often do you think someone meaningfully innovates and changes meaningfully our understanding of the world around us, and it's fundamentals? Consider, like, classical logic. It's pretty much done. You have the universal gate sets, and any novel functions are just going to be an mxn-mux. Most other such general concepts are essentially figured out and have had their applications recorded, explained, demonstrated. Sure, biographical information and history changes on a whim, but no one is requesting artificial clairvoyant intelligence. I suppose you could have the model pontificate and assess the truth of its own argumentation (ask it to generate outer-product concepts, ask it to reassess assumptions in current models) but it will quickly learn its own language that will be mutually unintelligible with the corpus of current human knowledge it has been trained on. Why does the underlying model need to change on a whim, in your opinion?... on what whim?

meltyness 4y ago

I like it for some stuff. The UI is certainly faster than ChatGPT, and seems to be more reliable at this stage. I suspect that its training has benefited from the unique corpus access relationships Google has accumulated since it's pre-IPO heyday. Think of the data connections they were establishing around 2005 with the establishment of Google Video and Google Books.

paulpauper 4y ago

Maybe this is good news for people who fear AI is becoming too powerful. Google is attracting talent that may otherwise go to Open AI, but diverted to produce something useless/harmless.

trash3 4y ago
[dead]

chrsstrm 4y ago

Bard is disappointing because it is incomplete. You can ask what it can do (sending mail, SMS, etc. and outputting formats like audio, video, JSON, CSV, etc) and it will happily tell you, but then prompt it to perform one of those actions and it just says Sorry, I can't do that yet. Maybe it will get better in time. The one product I'm absolutely disappointed in is Adobe Firefly. I just got access a couple days ago and this is the most alpha product wrapped in pretty ribbons I've seen in a while. All the cool features they teased like text-to-vector and sketch-to-image/vector are not available. In fact, only 2 out of the listed 12 features work, and while OK, they are not better than other options. Also take into account there's no prompt history so good luck getting back to an image set you liked and the biggest head scratcher is you can't use it for commercial purposes. This is a design product. From Adobe. And you can't use it for commercial purposes... Who would ever pay for this?

vitorgrs 4y ago
I mean, Adobe Firefly is free right now. Why would you pay? I found it Firefly quite good for food, portraits...

rhaway84773 4y ago

This could arguably be because Google is a lot better than ChatGPT right now. I asked chatGPT to create a meal plan, which it did, and all the recipes looked great, except, clicking the links for the recipes showed that they were completely made up, and either the links didn’t exist or were generated links, which in case they did were recipes for something completely different. Doing the same search at Google may have required a couple of clicks, but all the top listing would return a weekly meal plan.

vitorgrs 4y ago
For this use, you should.... use Bing, PHIND, etc which is GPT integrated to internet? Why would you compare a "offline" service to an online one when you want online things??
mustacheemperor 4y ago
For what it’s worth, I’ve found ChatGPT to be an effective weekly meal plan tool by just asking it to generate the recipes. It’s a word generation machine, and it can’t currently browse the live internet. I think it’s more suited to its design to ask it to create all the output than to summon accurate links from its knowledge set. If you want links to recipes by other people, I think Bing Chat would be more useful for you, since it can search the internet. There will eventually be a browsing plugin for ChatGPT, but it’s not available yet.
thrthrthr123 4y ago
> This could arguably be because Google is a lot better than ChatGPT right now. Lol. Nice try Mr Pichai I haven't used Google in more than a month now.
theRealMe 4y ago
Why do you need or want links for a recipe? ChatGPT(ESPECIALLY with gpt4) will tell you the full recipes in a nice concise easy-to-follow list. Why are you grading a chatbot on its web search result abilities instead of its answer-providing abilities? It’s like saying that horses are better than cars because the cars hurt when you sit on top with a leg down on each side.

jimsimmons 4y ago

[flagged]

thinkingemote 4y ago

Bard isn't bad objectivly it's just like GTP a year ago and we will see it improve quickly. It has more potential if it can work with the siloed tech and knowledge stores to be great. But it's a big if and might be too much for the org to handle. Remember Google still has all the world's books digitised. Maybe the copyright laws that prevent them using it normally will allow it's use in a LLM? Just like every other model uses copyrighted works.

llimos 4y ago

This article is missing the point. The story here is not that Google chose business over safety. It's that Google's AI is crap and they can't make it work well even if they want to.