I think this would only be acceptable if the “AI-assisted” system kicks in when call volumes are high (when dispatchers are overburdened with calls).

For anyone that’s been in a situation where you’re frantically trying to get ahold of 911, and you have to make 10 calls to do so, a system like this would have been really useful to help relieve whatever call volumes situation was going on at the time. At least in my experience it didn’t matter too much because the guy had already been dead for a bit.

And for those of you who are dispatchers, I get it, it can be frustrating to get 911 calls all the time for the most ridiculous of reasons, but still I think it would be best if a system like this only kicks in when necessary.

Being able to talk to a human right away is way better than essentially being asked to “press 1 if this is really an emergency, press 2 if this is not an emergency”.

I had to click to figure out just what an “AI Browser” is.

It’s basically Copilot/Recall but only for your browser. If the models are run locally, the information is protected, and none of that information is transmitted, then I don’t see a problem with this (although they would have to prove it with being open source). But, as it is, this just looks like a browser with major privacy/security flaws.

At launch, Dia’s core feature is its AI assistant, which you can invoke at any time. It’s not just a chatbot floating on top of your browser, but rather a context-aware assistant that sees your tabs, your open sessions, and your digital patterns. You can use it to summarize web pages, compare info across tabs, draft emails based on your writing style, or even reference past searches.

Reading into it a bit more:

Agrawal is also careful to note that all your data is stored and encrypted on your computer. “Whenever stuff is sent up to our service for processing,” he says, “it stays up there for milliseconds and then it’s wiped.” Arc has had a few security issues over time, and Agrawal says repeatedly that privacy and security have been core to Dia’s development from the very beginning. Over time, he hopes almost everything in Dia can happen locally.

Yeah, the part about sending my data of everything appearing on my browser window (passwords, banking, etc.) to some other computer for processing makes the other assurances worthless. At least they have plans to get everything running locally, but this is a hard pass for me.

I didn’t factor in mobile power usage as much in the equation before because it’s fairly negligible. However, I downloaded an app to track my phone’s energy use just for fun.

A mobile user browsing the fediverse would be using electricity around a rate of ~1 Watt (depends on the phone of course and if you’re using WiFi or LTE, etc.).

For a mobile user on WiFi:
In the 16 seconds that a desktop user has to burn through the energy to match those 2 prompts to chatGPT, that same mobile user would only use up ~0.00444 Wh.

Looking at it another way, a mobile user could browse the fediverse for 18min before they match the 0.3 Wh that a single prompt to ChatGPT would use.

For a mobile user on LTE:
With Voyager I was getting a rate of ~2 Watts.
With a browser I was getting a rate of ~4 Watts.

So to match the power for a single prompt to chatGPT you could browse the fediverse on Voyager for ~9 minutes, or using a browser for ~4.5 minutes.

I’m not sure how accurate this app is, and I didn’t test extensively to really nail down exact values, but those numbers sound about right.

Yeah, if you’re relying on them to be right about anything, you’re using it wrong.

A fine tuned model will go a lot further if you’re looking for something specific, but they mostly excel with summarizing text or brainstorming ideas.

For instance, if you’re a Dungeon Master in D&D and the group goes off script, you can quickly generate the back story of some random character that you didn’t expect the players to do a deep dive on.

Yeah, ~100-133 depending on how much energy your electric kettle uses.

Depends on the electric kettle, the first few I looked at on amazon run at ~600-800 Watts.

So, on the lower end there, you’re looking at about 0.166 Wh every second.

So a single push to chatGPT (0.3 Wh) uses about the same energy as an electric kettle does in less than 2 seconds.

While I agree that their comment didn’t add much to the discussion, it’s possible that you used more electricity to type out your response than it did for them to post theirs.

It’s estimated that a single ChatGPT prompt uses up ~0.3 Wh of electricity.

If @Empricorn@feddit.nl is on a desktop computer browsing the internet using electricity at a rate of ~150 W, and @TropicalDingdong@lemmy.world is on a smartphone, then you would only have ~16 seconds to type up a response before you begin using more electricity than they did.

I think you missed the part at the very end of the page that showed the timeline of them reporting the vulnerability back in April, being rewarded for finding the vulnerability, the vulnerability being patched in May, and being allowed to publicize the vulnerability as of today.

“environmentally damaging”
I see a lot of users on here saying this when talking about any use case for AI without actually doing any sort of comparison.

In some cases, AI absolutely uses more energy than an alternative, but you really need to break it down and it’s not a simple thing to apply to every case.

For instance: using an AI visual detection model hooked up to a camera to detect when rain droplets are hitting the windshield of a car. A completely wasteful example. In comparison you could just use a small laser that pulses every now and then and measures the diffraction to tell when water is on the windshield. The laser uses far less electricity and has been working just fine as they are currently used today.

Compare that to enabling DLSS in a video game where NVIDIA uses multiple AI models to improve performance. As long as you cap the framerates, the additional frame generation, upscaling, etc. will actually conserve electricity as your hardware is no longer working as hard to process and render the graphics (especially if you’re playing on a 4k monitor).

Looking at Wikipedia’s use case, how long would it take for users to go through and create a summary or a “simple.wikipedia” page for every article? How much electricity would that use? Compare that to running everything through an LLM once and quickly generating a summary (which is a use case where LLMs actually excel at). It’s honestly not that simple either because we would also have to consider how often these summaries are being regenerated. Is it every time someone makes a minor edit to a page? Is it every few days/weeks after multiple edits have been made? Etc.

Then you also have to consider, even if a particular use case uses more electricity, does it actually save time? And is the time saved worth the extra cost in electricity? And how was that electricity generated anyway? Was it generated using solar, coal, gas, wind, nuclear, hydro, or geothermal means?

Edit: typo

I really like the “Feeds” feature of PieFed, it allows anyone to combine all of the different communities into a single feed. It really makes browsing something like “Technology” a lot better.

Based on the uptick in “I was banned from Reddit” posts, I’m thinking that we’re getting a lot more users that were banned for good reason from Reddit. Looks like Reddit has also stepped up their game in their ability to keep those users off their platform.

It’s not sending the audio to an unknown server. It’s all local. From the article:

The system then translates the speech and maintains the expressive qualities and volume of each speaker’s voice while running on a device, such mobile devices with an Apple M2 chip like laptops and Apple Vision Pro. (The team avoided using cloud computing because of the privacy concerns with voice cloning.)

Highlighting the main issue here (from the article):

“This means that it is possible for the WhatsApp server to add new members to a group,” Martin R. Albrecht, a researcher at King’s College in London, wrote in an email. “A correct client—like the official clients—will display this change but will not prevent it. Thus, any group chat that does not verify who has been added to the chat can potentially have their messages read.”

Looks like the research was only looking at Denmark:

economists Anders Humlum and Emilie Vestergaard looked at the labor market impact of AI chatbots on 11 occupations, covering 25,000 workers and 7,000 workplaces in Denmark in 2023 and 2024.

Its probably better this way.

Otherwise you end up with people accusing movies of using AI when they didn’t.

And then there’s the question of how you decide where to draw the line for what’s considered AI as well as how much of it was used to help with the end result.

Did you use AI for storyboarding, but no diffusion tools were used in the end product?

Did one of the writers use ChatGPT for brainstorming some ideas but nothing was copy/pasted from directly?

Did they use a speech to text model to help create the subtitles in different languages, but then double checked all the work with translators?

Etc.

Too far would be anything outside of fair use. If a user generates an image of a specific copyrighted character, then attempts to make money off of that image, they could be sued.

You can’t copyright a style, but there’s still a lot of legal grey area here.

It’s also worth noting that OpenAI has an indemnification clause in their Terms of Use. This means that if someone else goes after OpenAI for something that went viral and was created by a specific user, OpenAI can then turn around and bill that user for all legal fees incurred by them (whether they win or lose the case).

If anyone is into using AI for anything, I would strongly suggest that they avoid using (or at least publishing/posting about) any of OpenAI’s tools especially while all of these legal issues are still being sorted out.