mztwo

5 ms·

1.
On-Device Acceleration of Large Diffusion Models(arxiv.org)
9 points by mztwo 3y ago|3 comments
2.
by mztwo 3y ago
What's so shocking in the brashness of their claims: - UMG alleges that training generative AI on any of their artists is a "breach of our agreements and a violation of copyright law." - We are asked, which side of history to
3.
by mztwo 3y ago
As GPUs like Nvidia's H100 cost over $30,000, Microsoft's own AI chip appears to arrive at a very opportune time to enable the company to pursue a very aggressive AI strategy.
4.
by mztwo 3y ago
The sheer size and scope of a single scam is absolutely astounding here -- and who knows how many others have yet to be discovered. When I search for new generative AI tools, it really does start to feel like a Wild West out there.
5.
by mztwo 3y ago
From a pure numbers perspective, it's no question that it's a good rush. I think Sam Lessin's perspective that the incumbents may reap most of the rewards here is interesting to consider --- some of the dynamics of the genera
6.
by mztwo 3y ago
What does a post intellectual property world even look like, though? I'm not aware of any convincing frameworks that our existing society could merge into.
7.
by mztwo 3y ago
From reading Reddit and seeing how people are dealing with Bing Chat's embedding of sources, it sounds like there is a lot of unanswered anxiety around what will happen to the internet if anything you put out there simply gets regurgit
8.
EU's AI Act: ChatGPT must disclose use of copyrighted training data(artisana.ai)
51 points by mztwo 3y ago|65 comments
9.
by mztwo 3y ago
The AI Act has been under development since 2021 (it's the EU... so it takes time) -- but news broke this week that there are additional provisions under discussion specifically designed to address the rise of chatbots. My full summary
10.
Assigning Personas Creates a Sixfold Increase in ChatGPT Toxicity(artisana.ai)
3 points by mztwo 3y ago|1 comments
11.
by mztwo 3y ago
Some notable things that stood out in the article are: Even baseline responses from a "good" persona would rate quite high on the toxicity scale the researchers used. There seems to be broad bias from all personas (and no-persona
12.
by mztwo 3y ago
I echo with your take. There's a lot of uncertainty, and in the meantime there's change that can be painful for some groups of professionals.
13.
by mztwo 3y ago
As the news media and twitterverse peddled the whole "crime-ridden SF" angle, I felt myself getting drawn into the same topic and agreeing with numerous friends on how much the city has declined. This is a good reminder of how we
14.
by mztwo 3y ago
Google is where projects go to die. Several friends who have joined via M&A have shared they had little expectations of seeing their babies continue in any way, shape of form after purchase by Google. While I don't know this founde
15.
by mztwo 3y ago
Source for this is the Run of World article linked in the document. Their China journalist spoke with recruiters who work on video game illustrator roles and gave background info on the trend and decrease. I'd like to think this is sim
16.
by mztwo 3y ago
Absolutely stunning how quickly this has disrupted an entire profession.
17.
by mztwo 3y ago
Lots of immediate business from companies needing humans to spin up their models though... but as LLMs get more advanced it's anyone's guess what will happen here.
18.
by mztwo 3y ago
Yep - you highlighted exactly what raised my eyebrow as I was writing the article.
19.
by mztwo 3y ago
To add to that, the leaked documents from Snowden described a query language not unlike doing Boolean searches. Nothing close to GPT’s ability to comprehend human asks.
20.
by mztwo 3y ago
Anecdotally, several CTOs I know intend to lessen their use of Scale, Labelbox and more in the future. Talked to one today who already ditched MTurk for GPT-4 -- cheaper, better, faster was what he said. Labelbox does image annotating still
21.
GPT-4 Outperforms Elite Crowdworkers, Saving Researchers $500k and 20k hours(artisana.ai)
147 points by mztwo 3y ago|144 comments
22.
by mztwo 3y ago
Buried in an arXiv paper was this nugget. Thought I'd share!
23.
by mztwo 3y ago
I would say the study itself is a groundbreaking milestone in the architecture it posits. The human behavior... quite mundane I agree! I watched the full demo twice and it reminded me of the more boring parts of the Sims 4. But maybe that&#
24.
by mztwo 3y ago
One outcome of this study was that a panel of evaluators judged the bot interactions to be more "human" than when humans impersonated these characters. So you have a point.
25.
by mztwo 3y ago
The paper explains that they used a panel of evaluators to judge the "humanness" of the interactions : )
26.
by mztwo 3y ago
Curious -- where do you think the article oversells the research paper? In reading through the full study a few times, what stood out to me was the impression these Generative Agents left on the authors -- despite having mostly mundane inte
27.
by mztwo 3y ago
What the researchers bolted on is an architecture that enables the storage and recall of memories, as well as self-reflection and more. They call out early the paper that even standard ChatGPT is not quite capable of this. ChatGPT here is u
28.
by mztwo 3y ago
The authors of the study were clear to call out there are quite a few downsides to mass adoption of Generative Agents... the pollution and misinformation angle certainly being top of mind. I'm inclined to agree.
29.
by mztwo 3y ago
The authors used gpt-3.5-turbo which does not like to spew adult content.
30.
Stanford's Groundbreaking AI Study Simulates Authentic Human Behavior(artisana.ai)
59 points by mztwo 3y ago|3 comments

More ›