Even Google Insiders Are Questioning Bard AI Chatbot's Usefulness – Slashdot

Become a fan of Slashdot on Facebook
Nickname:
Password:
Nickname:
Password:
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
Young Google took over the search world by storm and stole the crown of the large incumbent search engines of the early 2000s. It’s their time to have their lunch eaten by a challenger – in this case, OpenAI and their sugar daddy Microsoft.
google was the first to really up the game when it came compensating engineering talent – the 80’s IT giants were all management-first.
That attracted a lot of talent at first, but then the greedy came.
A new, prototype product doesn’t meet end-user expectations vs. another that’s been on the market longer.
The only advantage I’ve seen Bard vs. ChatGPT is that Bard does pull data from the Internet to support answers but overall the response queries aren’t as refined
as ChatGPT.
And then, what’s the point? OpenAI’s selling point is that their models are *really good* compared to open models. If you don’t have something that blows away free open models, then why would anyone care about your closed model?
Anyway, this article is kind of silly without knowing what percentage of the people in the chat had these sorts of views / made these sorts of comments. Makes for a nice “gotcha”, but not very good journalism.
The various AI tools have been wonderful at expanding my toolbox and offering alternative approaches to technical problems I have to solve day to day.
They are absolutely awful at generating accurate code, but they do give you an idea of what the solution would look like.
Ironically, the reason AI has made me a better developer is that it has offered me solutions, given me the wrong code, and then I put in the work to fix that code, and ended up learning more.
And I guess that’s something.

My rule of thumb is not to trust LLM output unless I can independently verify it
My rule is to not trust anything Google makes unless it’s open-source and reviewed. This rules out using Bard, because even if it was 100% accurate, it would probably he busy “monetizing” me behind my back and compromising my privacy in some way or other.
When it comes to coding, apparently my experience is a lot different from yours. ChatGPT has been a huge help in many situations, particularly where I’m using a language or technology I’m not that familiar with. No, it won’t write everything for me, but that’s not what I expect from it. It definitely has been a timesaver!
I agree with you about what’s taught (or not taught) in school, it doesn’t really prepare students to be engineers.
As far as working in large projects, yeah, I get it. I’ve been doing software development professionally for 35 years, in 28 different languages. These days, any one project requires doing work in half a dozen languages. Nobody knows every nuance of all of those languages, but a good developer knows what to expect from a language, but might not know exactly how to type the syntax to get what th
By total coincidence today, I thought I’d try giving Bard a bunch of customer feeback about something (about 500 free form responses) to see if it could summarise it.
First I asked: “Can Bard summarise customer feedback?” It said:
“Yes, Bard can summarize customer feedback. Bard is a large language model from Google AI, trained on a massive dataset of text and code. Bard can generate text, translate languages, write different kinds of creative content, and answer your questions in an informative way.
To summar

WTF??
Why the question marks? You typed and it engaged you with long detailed responsed that “felt” like they could be real. LLMs are specifically built to generate text that feels natural, and engages with the human prompter – and that’s what it did.
Good point. Well made.

WTF??
Why the question marks? You typed and it engaged you with long detailed responsed that “felt” like they could be real. LLMs are specifically built to generate text that feels natural, and engages with the human prompter – and that’s what it did.
Nah. To be interpreted as actual proper speech, there are some rules [wikipedia.org] it should generally follow. There are exceptions–when you are at odds with someone you can’t demand truthfulness–but if the chatbot talks about things that don’t exist, espousing capabilities it doesn’t have, it is not conversing properly.
(If you want to argue about the interface being separate from the backend, I don’t find that particularly compelling–the interface is always separate from the backend, including in humans. And you woul
Mind you, ChatGPT says…
Me:
I have a spreadsheet of about 500 lines of customer feedback. Can you generate a summary of that for me?
ChatGPT:
Certainly! If you could provide a few sample entries or key themes from your spreadsheet, I can help create a summary for you. You can share the general sentiments, common issues, or any specific aspects you’d like to focus on. This will enable me to generate a concise summary of the customer feedback without the need to disclose the entire content of the spreadsheet.
Me
First, extreme and overoptimistic hype, with pundits extrapolating well beyond actual capabilities
Then, disappointment as reality intrudes
Hopefully, it will be followed by genuine usefulness, but it’s not guaranteed
See VR for another example
Bard has been a disappointment from the start. To such an extent that it actually hit google stock when it was first presented in the wake of ChatGPT 3 unveiling to the masses, and making errors in the presentation where it was unveiled.
Indeed. Actual usefulness with LLMs will be limited though, because you cannot stop them from hallucinating unless you make them extremely limited.

My rule of thumb is not to trust LLM output unless I can independently verify it.
This is a good mantra … not only for LLMs but for everything one finds on the internet and elsewhere. Imagine how much better our world would be if this mantra were applied to Google search, social media, extremist websites, and even Wikipedia. ChatGPT, Bard, etc. are getting a bad rap because they’re being used/held wrong. LLMs are a fantastic way to take a first step at exploring a topic, just like Google search and Wikipedia. Copying and pasting Google search or Wikipedia content as a finished solution is foolish. Anyone who uses LLMs in a similar way is just as foolish.
LLMs are held to that standard because they make mistakes the most stupid rookie would generally not make.
The future is looking more open-source anyway.
Companies will likely continue to struggle monetizing the tech due to how open-source it is. The ones that are trying to do so, with their subscriptions and advertising models are still hemorrhaging money. And that is before people decide that the free versions are just good enough, if not better then the paid versions. Or if they really want to, just train their own model for whatever purpose they need.
So perhaps you are right that AI tech is here to stay. But
I guess you have not been around for the last AI hype. No, they _cannot_ make it work. What they currently have is the best that this tech can deliver.
I guess you have never seen an AI hype before. This is not the first time all those things you expect have been promised. It will not work out this time either.
Indeed. Same here. “AI” proponents has been some of the greatest purveyors of false advertising and exaggerated claims of all time. The only thing that can compete is religion and some politicians. And the AI people do it again and again.
IBM Watson flashback!
Highly disappointing.
You really do not get it. Hence you are “shocked” at something quite obvious. The thing is, LLMs are testet for _replaciong_ experts, not for making things a bit easier for an actual expert.
That may be the most stupid and insightless comment here. Yes, you test. But you never test fully and you do not expect really, really stupid but non-obvious mistakes. And you cannot really test code security. You have to get that right by design, there is just no other option.
Well, Google was flashy and elite some time ago. Now they seem to have their fair share of idiots.
There may be more comments in this discussion. Without JavaScript enabled, you might want to turn on Classic Discussion System in your preferences instead.
Microsoft Gives Unexpected Tutorial on How To Install Linux
Adobe Previews AI Upscaling To Make Old, Fuzzy Videos and GIFs Look Fresh
The biggest difference between time and space is that you can’t reuse time. — Merrick Furst

source