I guess what I don't understand, why does AI have to look like magic omnipotence? We put in a question, and this software spits out an answer based on the data it's leeched off of other sources.
Why isn't there an expectation that an AI will cite it's sources like a wiki article?
To call 'prompt engineering' engineering is :'(. No first principles on which to base 'prompt designs'. No guarantees that a prompt would work as expected in all cases/expected cases.
As with everything else it has touched, #LLM#tech has redefined 'engineering'.
Terming this 'engineering' has real impact. Young scientists/engineers believe that all of this is 'proven' - just like in software engineering.
Oh, that is beautiful! A tool to confuse AI models stealing your copyrighted images. It is called glaze. You run your pictures through it before uploading somewhere. That is a start until laws are enforced.
I like to listen to childhood nostalgia, and 1980s Italo Disco is among that.
There's a 1985 song by Den Harrow called Future Brain, and I always had to grin listening to the naive song text, but with the current AI developments, that song text is starting to get a new, retro-prophetic meaning.
Do you work with #voice or #speech#data? You might contribute data, write data specifications for collection, perform filtering or pre-processing, train #ASR or #TTS models, or design or perform evaluations on #ML speech models.
If so, I’d love your help to understand current #dataset#documentation practices, and what we can do to make them better as part of my #PhD#research
The #survey takes 10-20 minutes to complete, and you can opt in to win one of 3 gift cards valued at $AUD 50 each
Research Protocol 2021/427 approved by #ANU Human Research Ethics Committee
I'm Jin (he/him), a staff data scientist and occasional #ml engineer in #fintech, and #compsci researcher specialising in hashing and anomaly detection, with an interest in graph networks.
Do you work with #voice or #speech#data? You might contribute data, write data specifications for collection, perform filtering or pre-processing, train #ASR or #TTS models, or design or perform evaluations on #ML speech models.
If so, I’d love your help to understand current #dataset#documentation practices, and what we can do to make them better as part of my #PhD#research
The #survey takes 10-20 minutes to complete, and you can opt in to win one of 3 gift cards valued at $AUD 50 each.
Research Protocol 2021/427 approved by #ANU Human Research Ethics Committee
So should we put some kind of magic moratorium on ML research for a few months as some technologists have self-servingly suggested? No, that's just silly. What we need to do is recognize these risks and face them directly."
"This video depicts a fictional #podcast between Joe Rogan and Sam Altman, with all content generated using AI language models... The use of AI technology to generate this content is solely intended as an exploration of the capabilities of language models and should not be misconstrued as a genuine conversation between the individuals depicted... Viewers are encouraged to approach this content with a critical and discerning eye..."
Well, that’s what I call #progress. Ported my existing CameraView using #Camera2 on #Android code to #MAUI. Next step=> use Google’s #ML#Vision to read Barcodes and also recognize Text.
Wondering what kind of weather you are going to have today? Add a Weather Widget to your Desktop or Panel to be in the know.
See the guide at: https://kfocus.org/wf/weather
#KubuntuFocus strives to make work easier and more enjoyable.
#AR is coming. Great! #XAI is important. Of course!
But how to combine them together?
Our paper "XAIR: A Framework of Explainable AI in Augmented Reality" has won an Honorable Mention award #CHI2023! Credits to the team at Meta Reality Labs in the summer! #AI#ML
Large language models (LLMs) are becoming increasingly popular and more widely accessible for commercial use, as we can see with the rise of OpenAI's #ChatGPT and Google's #LaMDA.
These #ML models are able to generate text that can be indistinguishable from text written by a human, and can be used for a wide range of #NLP tasks such as translation, text summarization, and answering questions.
But these LLMs invariably come with #privacy implications that must be addressed.
Patricia Thaine (Co-Founder & CEO, Private AI) explores how LLMs are being used (and misused), what risks companies leveraging these technologies should be thinking about in respect to exposing #PII, and what solutions exist to protect their customers’ personal #data.
From Adrian Roselli: AI-Generated Images from AI-Generated Alt Text
My thoughts: Yeah I sat there and was like "Wow, image recognition is really good." And then I've read posts like this and I'm like yeah AI gives a loose idea, but it isn't going to replace caring, helpful people allowing us to appreciate their images, art, and so on any time soon. And he got hit with the way Apple's ML just goes "we don't recognize this so we'll call it a video game." As a blind person, ML generated descriptions are like a person who slept through a book summarizing it in 10 words. When an image is described by a person, it's more like experiencing the thing. And while I'll still use ML for when nothing else is available, it really shows the difference between "well it has Alt Text '' as opposed to "this has an excellent description that shows that the uploader of the image has a higher than average level of human decency.
If they can start getting their drivers right, they might be able to compete in the consumer market against nVidia, AMD and Intel. That's if they choose to go that route. The #AI, #ML and #Crypto markets might be enough for them without going after the #consumer market.
Either way, it seems that the age of the Chinese GPU is coming.