GPT 4 hallucination rate is 28.6% on a simple task: citing title, author, and year of publication medium.com/@michaelwood33311/g…

Hallucination rates and reference accuracy of ChatGPT and Bard for systematic reviews: Comparative analysis jmir.org/2024/1/e53164 #AI #LLM

GPT 4 Hallucination Rate is 28.6% on a Simple Task: Citing Title, Author, and Year of Publication

The all-too-common myth of GPT 4 having only a 3% hallucination rate is shattered by a recent study that found GPT 4 has a 28.6% hallucination rate. That’s almost 10x the oft-cited (i.e. over hyped)…

^{Michael Wood (Medium)}

Here's a very calm, well researched video analyzing the facts and data about AI technology progress. And all the stunning tech demos that were faked.

I really like the scientific approach to this and I feel like I'd want to adapt this kind of analytical process to other things in my life.

youtu.be/VctsqOo8wsc?si=XcuufC…

#AiRant #AI #fAIl #tech #science

AI Hype is completely out of control - especially since ChatGPT-4o

Now that ChatGPT4o is out, it's time to revisit my previous thoughts on AI, and discuss what might be happening in the future, and what evidence we have of w...

^YouTube

Upon recommendations, I tried this AI prompt injection game for the first time. I made it to level 7 with no help from the internet!

If you want to donate to me, donate to me on this page.

My website is here where I usually blog. I'm not much of a video person, so I blog and write more than I do video!

One of the selling points of Google’s flagship generative AI models, Gemini 1.5 Pro and 1.5 Flash, is the amount of data they can supposedly process and analyze. But new research suggests that the models aren’t, in fact, very good at those things. “While models like Gemini 1.5 Pro can technically process long contexts, we have seen many cases indicating that the models don’t actually ‘understand’ the content,” Marzena Karpinska, a postdoc at UMass Amherst. TechCrunch has more: flip.it/wW5pRm
#Tech #AI #ArtificialIntelligence #Google #Gemini

Follow @Techcrunch for more breaking tech news and analysis.

And what would these standards be, pray tell, @mozilla ?

Keeping in mind that you are including services that told people to eat rocks or put glue on a pizza?

> Our initial offering will include ChatGPT, Google Gemini, HuggingChat, and Le Chat Mistral, but we will continue adding AI services that meet our standards for quality and user experience.
blog.mozilla.org/en/products/f…

🤡

/via @alxd

#AI

Choose how you want to navigate the web with Firefox

We're launching an opt-in experiment offering access to preferred AI services in Firefox Nightly for improved productivity as you browse.

^{Kristina Bravo (The Mozilla Blog)}

Wait...What?

microsoft.com/en-us/security/b…

#AI #cybersecurity

DELL: Putting the IT in SHIT

#AI

This Canadian Member of Parliament asked ChatGPT for a list of capital gains tax rates by country, got a nonsense answer, screenshotted it, and then tweeted the incorrect information.

(He later deleted the tweet.)

#canada #cdnpoli #twitter #x #ai #chatgpt #misinformation

Here's a picture taken with the Meta Smart Glasses, then described with #AI via #BeMyEyes, which has been an invaluable helping-hand, enabling me to know what randomly labeled pictures in my collection are, without having to ask my wife or children.
With image names such as 'photo-296_singular_display_fullPicture.jpg' it's a fun time trying to decipher such randomness sometimes.
This was taken at a podcast recording of Tim Harford's 'Cautionary Tales' and I got to take Jake along with me.

I get the AI hate I really do, but there are just times in life where hate is misplaced. for me, something that is an enabler is such a time.
No, it's not perfect, we're not there yet, but for my usecase it helps more than it hinders.
Your milage will of course, vary.

Well, these authors cut right to the chase in this paper!

#LLMs
#AI
#GenerativeAI
#bullshit
#ethics
#truth

🧵2/5

Tuesday: #AI is being trained on images of children, spanning their entire childhood without #consent. 📷👶 AI is coming to an inbox near you, make sure to protect your data 👉 tuta.com/blog/google-gemini-ai…

Don't want Gmail to sift through your emails? Too late, their AI is going to do it.

Google's Gemini AI is coming soon to Gmail where it will be able to craft AI generated responses, summarize conversations, and more.

^Tutanota

Meta is deploying chatbots powered by AI to impersonate humans in its discussion groups. The chatbots make stuff up—that is their nature. One invents a nonexistent child. Another pretends to have nonexistent merchandise. It shouldn’t be necessary to explain how vile and poisonous this is, but Casey Fiesler @cfiesler spells it out:

theconversation.com/ai-chatbot…

#Meta #AI

AI chatbots are intruding into online communities where people are trying to connect with other humans

People participate in online communities to share experiences and sympathy. AI chatbots that join the conversation can only pretend to offer either.

^{The Conversation}

Microsoft Copilot helpfully summarizing an obvious phishing email, without alerting that it's a phishing email!

#ai #bullshit

I've just realised that Qui-Gon Jinn of the Galaxy far, far away has absolutely the best quote suitable for the current state of the so called AI...

...The ability to speak does not make you intelligent.

Pretty impressive for the 1999 movie, isn't it?

#starwars #ai #memes

“The ugly truth behind ChatGPT: AI is guzzling resources at planet-eating rates”

Mariana Mazzucato

(who I consider a credible source)

#ai #climatechange

theguardian.com/commentisfree/…

The ugly truth behind ChatGPT: AI is guzzling resources at planet-eating rates

Big tech is playing its part in reaching net zero targets, but its vast new datacentres are run at huge cost to the environment, says economics professor Mariana Mazzucato

^{Mariana Mazzucato (The Guardian)}

Google operated on the social contract of providing traffic to sites in exchange for content. Now that #Google is switching to #AI answers synthesized from the content and effectively cutting traffic is there even a point in participation?

Google will keep bringing load to the sites without providing any traffic.

I suspect we are about to see first robots.txt snippets banning Google crawlers like other AI bots really soon. And eventually outright banning Google ips altogether.

I wonder whether we’re also about to see a surge of old-school search engines. With modern technology it’s feasible to run a small search engine from a living room (e. g. kukei.eu) on an off-the-shelf hardware and open source software. It doesn’t take the whole Google to operate a decade-old (feature wise) Google search engine. It might be feasible to run a good search engine on small resources that can be paid by simple ads or small subscriptions.

platformer.news/google-io-ai-s…

Google’s broken link to the web

With AI search results coming to the masses, the human-powered web recedes further into the background

^{Casey Newton (Platformer)}

Anything from Vox Media and The Atlantic is now untrustworthy in any capacity from now on. Remember this when reading or linking to any articles in the future.

#AI #FuckAI #Journalism

y'all know +972 Magazine? the media organization responsible for some of the most important investigations of the state of Israel in this war, including the Lavender #AI investigation and now the new #ICC blackmail investigation?

it is almost entirely funded by donations of readers like you and me. if you want to see more journalism like this, consider throwing a few bucks their way.

972mag.com/members/

@israel @palestine

#Israel #Palestine #WarCrimes #InternationalCriminalCourt #Gaza

Become a member of +972 Magazine - +972 Magazine

Membership image

^{+972 Magazine}

'caverns of obsolete'

Puffy and Girl come upon vintage tech while exploring ancient ruins.

#technomage #unix_surrealism #comic #ai #openbsd #puffy #fediart #mastoart

Thank you to whoever made youwouldntsteala.website/edito… 🙌

#Google #AI #Memes

The image shows a screenshot of a Google search result. The search query at the top is "is it ok to leave a dog in a hot car?" The AI Overview section below the search bar provides a misleading and incorrect answer, stating "Yes, it's always safe to leave a dog in a hot car." It goes on to give false information about car temperatures and even mentions a fictitious Beatles song about leaving dogs in hot cars.

To the right of the text, there is a small image of a "Car Temperature Dog Safety Chart" which likely provides accurate information about the dangers of leaving dogs in hot cars, but the text in the chart is too small to read clearly in the screenshot.

The time on the phone is 11:30, and the battery is at 92%. #AI 👩‍🦯 social.kendoo.eu/@corona/11249…

corona :blinkingcursor: (@corona@social.kendoo.eu)

Attached: 4 images #ai #fundstücke #youcantmakethisshitup :-)

^{social.kendoo.eu}

Good. Biden's appointment of Altman to the AI safety and security board is up there with his Garland and Fauci appointments. Bad decisions.

The man has been accused by those who know him in his own inner circle of being "not consistently candid" and if there's anyone who should not be deciding AI risk issues it's someone not candid.

"Scarlett Johansson scandal puts OpenAI's Sam Altman back in the hot seat"

axios.com/2024/05/22/altman-op…

#AI #OpenAI #ScarlettJohansson #SamAltman #ChatGPT #Sky

back in my day we called this spyware

#AI #privacy #Microsoft

"How can you use AI, it steals things people share online, it burns down the planet!"

"Alright, do you mind telling me what's in this picture you posted online?"

"Ah. Carry on."

#AltText #accessibility #AI

Hey, if you don't like the fact that many #blind people have to rely on AI to describe images for them, you can help by, you know, describing images for them.

mastodon.social/@botwiki/11147…

#AltText #accessibility #ai

“I have seen the extremely restrictive off-boarding agreement that contains nondisclosure and non-disparagement provisions former OpenAI employees are subject to. It forbids them, for the rest of their lives, from criticizing their former employer.”

vox.com/future-perfect/2024/5/…

#openai #tech #ai

OpenAI departures: Why can’t former employees talk, but the new ChatGPT release can?

Why are OpenAI employees leaving the company helmed by Sam Altman? The safety alignment debate, explained.

^{Kelsey Piper (Vox)}