Google announces ChatGPT rival Bard, with wider availability in ‘coming weeks’

Mega6 said:
I would imagine if you're being replaced by chatgpt - you are not that technical to begin with.
Click to expand...
Google, ChatGPT, whatever comes next, there will always be a tool out there with the answer. The challenge is to be the tool that knows what is the right question, be that tool!
 
Lakados said:
Google, ChatGPT, whatever comes next, there will always be a tool out there with the answer. The challenge is to be the tool that knows what is the right question, be that tool!
Click to expand...
At the beginning of my career, I was using yahoo and then google to figure out my IT issues. Everyone else was still paging through books. 😉

Ya if chatgpt or whatever can help so be it. You have to learn to utilize it effectively, or just be replaced.
 
GoldenTiger said:
Concerned he's on target? Scared of people who have differing opinions?
Click to expand...
yes, foundationally shaken and not just marveling at a one track mind - clearly

Decko87 said:
I think the particular case he used it for initially is what it's best for, a simple, straightforward piece of code you'd see in a textbook. Once he got into a more advanced coding scenario he had to continually prompt it to get it right. That's a similar experience I had earlier but with something much simpler. It didn't understand that a Transform was an object and therefore not pass by value. It was trying to copy it into a variable to store the original location, scale, and rotation data, the only thing it actually did was copy a reference location. I corrected it, and it fixed the problem. The other issue was how inefficient the code was, but not a real complaint. My worry with my particular case, is a jr or a new programmer might not understand that concept or know that it's happening and struggle to understand why it's not working (it's a problem I ran into early in my coding career). The sieve he wrote is something you'd see in an algorithms class in college though, I did similar problems what I took an algorithms course online a few years ago (I'm self taught).
To be fair to ChatGPT, i'm pretty sure it wasn't trained on a ton of Unity and C# data so I'm sure it'll improve massively down the road.
Click to expand...
Amusingly, I've found the best thing about GitHub Copilot is how good it can be at documenting shit for you.

The actual code it generates more often than not is useless, unless it's stupid simple. And I've found half their promises don't really work like "it'll write tests for you!" or "just describe the function in a comment and it'll write it!"

No the fuck it won't unless you're implementing something incredibly cookie cutter.

Documentation however it rocks at. And better yet is it picks up on patterns. So you write something once and it'll model its next predictions off it. Which rules because it starts generating everything in a perfectly consistent formatting.

Also they're pretty dope for generating data sets. Give it the format and have it synthesize however many rows of just the kind of bullshit data you need in seconds.

On the other hand, I recently saw it comment something like "Day of the week" next to a variable named "Month" and it was a nice reminder that we don't yet need to fear Skynet
 
socK said:
yes, foundationally shaken and not just marveling at a one track mind - clearly


Amusingly, I've found the best thing about GitHub Copilot is how good it can be at documenting shit for you.

The actual code it generates more often than not is useless, unless it's stupid simple. And I've found half their promises don't really work like "it'll write tests for you!" or "just describe the function in a comment and it'll write it!"

No the fuck it won't unless you're implementing something incredibly cookie cutter.

Documentation however it rocks at. And better yet is it picks up on patterns. So you write something once and it'll model its next predictions off it. Which rules because it starts generating everything in a perfectly consistent formatting.

Also they're pretty dope for generating data sets. Give it the format and have it synthesize however many rows of just the kind of bullshit data you need in seconds.

On the other hand, I recently saw it comment something like "Day of the week" next to a variable named "Month" and it was a nice reminder that we don't yet need to fear Skynet
Click to expand...
I'll have to try it out again, I'm sure I can get my boss to buy me a subscription, I hate documenting my code but I do it.
 
When ever chatgpt can change out my tie rod ends, weld me a bracket, change out a start compositor or fix me a sandwich then I will worry.
 
There is a mcdonalds that is fully robotic, and cars are welded by robots, so it isnt a huuuuge stretch :p
 
mullet said:
When ever chatgpt can change out my tie rod ends, weld me a bracket, change out a start compositor or fix me a sandwich then I will worry.
Click to expand...
dude, that's going to be me tomorrow morning by 11am. Well, minus the start compositor- I don't even know what that is.
 
erek said:
this is compounded by the fact that the "Web" itself is a blurry image that is a blurry image of some other intermediary towards our own experience of reality which is probably some sample of reality






Message #general
Click to expand...
talking to yourself now?! ;)
i like this comparison

806489_1675997554508.png
 
Westwood said:
dude, that's going to be me tomorrow morning by 11am. Well, minus the start compositor- I don't even know what that is.
Click to expand...


If you know how to work with your hands and can fix things you are good to go, but if you sit behind a desk all day, well.......
 
mullet said:
If you know how to work with your hands and can fix things you are good to go, but if you sit behind a desk all day, well.......
Click to expand...
I do both. XD

Updated BIOS and installing a new CPU tonight. Tie rods, drag links, center link, and steering stabilizer tomorrow. Taxes and drywall on Sunday.
 
1_rick said:
Just the ordering side. The food is still cooked by people, but most people missed that part of the article.
Click to expand...
Fair. Was something I read on my phone so with the amount of ads and crap I tend to skip around
 
I'm watching ChatGPT generate "a C# gui program that accepts drag and drop 2 color bitmap files and crops any white border larger than one pixel on all four sides" right now, and it doesn't seem completely wrong, although it *gasp* uses goto repeatedly to break out of loops early, and it took "two color bitmaps" to mean two bitmap files, rather than two-bit color.

It also doesn't *save* the cropped bitmaps, which makes the program somewhat worthless, but I guess this shows how much effort you have to go to to get it to do what you want. (The version of the program I wrote will rename the original file and create a new file with the same name as the original.)
 
mullet said:
When ever chatgpt can change out my tie rod ends, weld me a bracket, change out a start compositor or fix me a sandwich then I will worry.
Click to expand...
The results from Mazda’s new precision welder. 1mm cube made of 0.05mm plates.
My Japanese is terrible so I can’t tell if that was done by a person but no names are mentioned which makes me believe it was a machine. Because if a member of their staff did that they would be bragging.
1676047194987.png
 
1_rick said:
I'm watching ChatGPT generate "a C# gui program that accepts drag and drop 2 color bitmap files and crops any white border larger than one pixel on all four sides" right now, and it doesn't seem completely wrong, although it *gasp* uses goto repeatedly to break out of loops early, and it took "two color bitmaps" to mean two bitmap files, rather than two-bit color.

It also doesn't *save* the cropped bitmaps, which makes the program somewhat worthless, but I guess this shows how much effort you have to go to to get it to do what you want. (The version of the program I wrote will rename the original file and create a new file with the same name as the original.)
Click to expand...
GoTo statements are funny things, they are super efficient but because we humans are chaotic and disorganized they become dangerous because we forget they are there.
 
Lakados said:
The results from Mazda’s new precision welder. 1mm cube made of 0.05mm plates.
My Japanese is terrible so I can’t tell if that was done by a person but no names are mentioned which makes me believe it was a machine. Because if a member of their staff did that they would be bragging.
View attachment 547984
Click to expand...
The US can also train grasshoppers to weld, I don't see the big deal.
 
Lakados said:
The results from Mazda’s new precision welder. 1mm cube made of 0.05mm plates.
My Japanese is terrible so I can’t tell if that was done by a person but no names are mentioned which makes me believe it was a machine. Because if a member of their staff did that they would be bragging.
View attachment 547984
Click to expand...
Weak, I’d be impressed with microscopic welds of this quality tho

1676047965444.jpeg


1676048034497.jpeg
 
1_rick said:
I'm watching ChatGPT generate "a C# gui program that accepts drag and drop 2 color bitmap files and crops any white border larger than one pixel on all four sides" right now, and it doesn't seem completely wrong, although it *gasp* uses goto repeatedly to break out of loops early, and it took "two color bitmaps" to mean two bitmap files, rather than two-bit color.

It also doesn't *save* the cropped bitmaps, which makes the program somewhat worthless, but I guess this shows how much effort you have to go to to get it to do what you want. (The version of the program I wrote will rename the original file and create a new file with the same name as the original.)
Click to expand...
Second iteration's version of "a 2-color bitmap" rejects files whose names don't end in ".bmp". FAIL. Acceptable options would be something like "count the number of distinct pixel colors, look at the image header to see if it specifies a 2-color format, or confirm there's a 2-entry color palette." Also, it assumes that the border is white, which isn't a bad assumption, but isn't strictly correct, but I now realize my hand-written program doesn't do that either.

I feel like getting it to work better than what I wrote would take at least as much time painstakingly tuning my description as it took me to write the program.

Oh, also, at the bottom was a note that the generated program assumed a Visual Studio project with a reference to System.Drawing already existed, and had assumptions based on that. Not wrong per se but precludes you from a "notepad and csc.exe" implementation.
 
1_rick said:
Second iteration's version of "a 2-color bitmap" rejects files whose names don't end in ".bmp". FAIL. Acceptable options would be something like "count the number of distinct pixel colors, look at the image header to see if it specifies a 2-color format, or confirm there's a 2-entry color palette." Also, it assumes that the border is white, which isn't a bad assumption, but isn't strictly correct, but I now realize my hand-written program doesn't do that either.

I feel like getting it to work better than what I wrote would take at least as much time painstakingly tuning my description as it took me to write the program.

Oh, also, at the bottom was a note that the generated program assumed a Visual Studio project with a reference to System.Drawing already existed, and had assumptions based on that. Not wrong per se but precludes you from a "notepad and csc.exe" implementation.
Click to expand...
Is this literally a 2 arbitrary color bitmap, or a 1-bit monochrome bitmap?

You said 2-bit at one point but that's 4 colors.
 
socK said:
Is this literally a 2 arbitrary color bitmap, or a 1-bit monochrome bitmap?

You said 2-bit at one point but that's 4 colors.
Click to expand...
well, 2-color, so 1-bit mono, but I just remembered that actually sometimes they're actually 8-bit color coming in, but I convert them to 4-bit palettized for file size reasons.

In the interest of fairness I just looked at my program and it also uses goto to exit out of the scanning loops, and I crop based on the color of the pixel in the top-left corner, which has always been good enough for me, but I guess "scan the perimeter and use the color with the most pixels as the crop color" would be more intelligent.
 
1_rick said:
well, 2-color, so 1-bit mono, but I just remembered that actually sometimes they're actually 8-bit color coming in, but I convert them to 4-bit palettized for file size reasons.

In the interest of fairness I just looked at my program and it also uses goto to exit out of the scanning loops, and I crop based on the color of the pixel in the top-left corner, which has always been good enough for me, but I guess "scan the perimeter and use the color with the most pixels as the crop color" would be more intelligent.
Click to expand...
"Large language models (LLMs) have shown impressive performance on complex reasoning by leveraging chain-of-thought (CoT) prompting to generate intermediate reasoning chains as the rationale to infer the answer. However, existing CoT studies are mostly isolated in the language modality with LLMs, where LLMs are hard to deploy. To elicit CoT reasoning in multimodality, a possible solution is to fine-tune small language models by fusing the vision and language features to perform CoT reasoning. The key challenge is that those language models tend to generate hallucinated reasoning chains that mislead the answer inference. To mitigate the effect of such mistakes, we propose Multimodal-CoT that incorporates vision features. The framework separates the rationale generation and answer inference into two stages. By incorporating the vision features in both stages, the model is able to generate effective rationales that contribute to answer inference. With Multimodal-CoT, our model under 1 billion parameters outperforms the previous state-of-theart LLM (GPT-3.5) by 16% (75.17%→91.68%) on the ScienceQA benchmark and even surpasses human performance. Code is publicly available.1"



https://www.marktechpost.com/2023/0...rms-gpt-3-5-by-16-75-17-→-91-68-on-scienceqa/
 
Google really got sidetracked out of AI into phones, gmail, quantum computing, ect. Dumb voice assistant that's flaky at best which I will never buy. . They indexed the internet and have continued to feed of of that. The basic google search engine really hasn't changed much in decades. Maybe it understands search phrases better, idk. Was always good at just giving it key words.
 
Mega6 said:
Google really got sidetracked out of AI into phones, gmail, quantum computing, ect. Dumb voice assistant that's flaky at best which I will never buy. . They indexed the internet and have continued to feed of of that. The basic google search engine really hasn't changed much in decades. Maybe it understands search phrases better, idk. Was always good at just giving it key words.
Click to expand...
Now they’re constantly exploited by serving up ad stuffing / stuffed malware clones of legitimate products in their key and sponsored search results. Theoretically google should be considered a security vulnerability at best
 
erek said:
Now they’re constantly exploited by serving up ad stuffing / stuffed malware clones of legitimate products in their key and sponsored search results. Theoretically google should be considered a security vulnerability at best
Click to expand...
Ya there is really no need for this other than google being lazy. It's beyond theoretical - it's fact.

EDIT - Personally, I've migrated to duck duck go for 90% of my searches now.
 
Last edited:
unfortunately, Google Home/Google Assistant is still the shiniest turd in the bucket in terms of smart home assistants, best device compatibility + features, they're all shit this just does more with more when it isn't compared to Alexa/Echo and Siri/HomeKit
 
staknhalo said:
unfortunately, Google Home/Google Assistant is still the shiniest turd in the bucket in terms of smart home assistants, best device compatibility + features, they're all shit this just does more with more when it isn't compared to Alexa/Echo and Siri/HomeKit
Click to expand...

My favourite thing about Siri is that it's let down in the end by Apple's demand that everything Siri does is somehow an in-house fruit product or service or whatever.

Mind you, the app store is what hamstrings the iPad.
 
staknhalo said:
unfortunately, Google Home/Google Assistant is still the shiniest turd in the bucket in terms of smart home assistants, best device compatibility + features, they're all shit this just does more with more when it isn't compared to Alexa/Echo and Siri/HomeKit
Click to expand...
Can it be polished into something killer?
 
erek said:
Can it be polished into something killer?
Click to expand...

It's main problem (that some say they all suffer from to varying degrees) is that it will recognize commands fine when you set it up, then 6 months later no longer recognize those same commands again when you say things in the exact same way/cadence/etc, then go back to working just fine again, then not, rinse and repeat - it's like they can't stop fucking with things on the backend or something

And recently it keeps resetting my assistant voice settings constantly from Australian lady to default voice
 
  • Like
Reactions: erek
like this
staknhalo said:
It's main problem (that some say they all suffer from to varying degrees) is that it will recognize commands fine when you set it up, then 6 months later no longer recognize those same commands again when you say things in the exact same way/cadence/etc, then go back to working just fine again, then not, rinse and repeat - it's like they can't stop fucking with things on the backend or something

And recently it keeps resetting my assistant voice settings constantly from Australian lady to default voice
Click to expand...
IBM’s Cloud Watson acted in a similar way for us. Set it up for some responses and it acted appropriately but then over night it “trained” itself and became unpredictable and essentially useless to rely on for even a simple demonstration. The Watson there doesn’t seem like the jeopardy Watson at all. It was far from impressive
 
Seems chatgpt and Barf 🤢 🤮 are the new Wikipedia in terms of scholastic relevance and acceptance. Yeah, a blurry image of the web is a partially accurate analogy
 
  • Like
Reactions: Mega6
like this
You must log in or register to reply here.
Back
Top