thorum

22 ms·

271.
by thorum 3y ago
What are you basing that on?
272.
by thorum 3y ago
> Image generation parameter count is hitting diminishing returns, going by what we've seen with SD/SDXL. For image quality, sure - language understanding is still an issue. SDXL can generate a beautiful image, but if it doesn’
273.
by thorum 3y ago
When we reach the point when average people can easily run today’s models on their own devices, today’s models will no longer be SOTA and there will still be demand to run better models in the cloud.
274.
by thorum 3y ago
It doesn’t seem to detect subdomains set up with Kubernetes ingresses, based on results for one of my domains, so that might be a place to start research.
275.
by thorum 3y ago
Threads was a breath of fresh air at first and I followed a bunch of people I already followed on Instagram and Twitter. Opening the app today there is exactly 1 post from someone I follow and pages upon pages of ‘recommended’ content from
276.
by thorum 3y ago
Honestly seems like the market itself should have safeguards against this kind of thing.
277.
by thorum 3y ago
5G surgery has been done, though: https://www.huawei.com/en/huaweitech/industry-insights/outlo...
278.
by thorum 3y ago
The author doesn’t appear to consider the idea that an LLM can be placed into a larger system where it isn’t limited to a single continuous inference, and given the ability to self-direct, i.e. the output of one inference can be the decisio
279.
by thorum 3y ago
replit’s model is surprisingly good at generating code, even at following complex instructions that I was sure would confuse it. I have found it’s a bit weak on code analysis , for open-ended questions like ‘is there a bug anywhere in this
280.
by thorum 3y ago
In a response about the Turing test on this site, LLaMa 2 used the phrase “to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human” which appears to be copied verbatim from the first sentence of the Wikiped
281.
OpenAI Worries About What Its Chatbot Will Say About People’s Faces(nytimes.com)
7 points by thorum 3y ago|1 comments
282.
by thorum 3y ago
It’s cheaper to use OpenAI. If you have your own compute, sentence-transformers is just as good for most use cases.
283.
by thorum 3y ago
I actually think this feature would be fine if implemented better. GPT-4 is perfectly capable of accurately explaining what a short code block does if you prompt it with relevant documentation… which MDN has in abundance, but isn’t using, f
284.
Twitch Streamer XQC Moves to Kick in $100 million deal(forbes.com)
4 points by thorum 3y ago|0 comments
285.
by thorum 3y ago
The JSON schema not counting toward token usage is huge, that will really help reduce costs.
286.
by thorum 3y ago
The subreddit in question is still private: https://old.reddit.com/r/tumblr
287.
by thorum 3y ago
“Melody conditioning” as shown in the article seems both immediately useful and something that’s harder to find a human to do for you at the same level of quality.
288.
by thorum 3y ago
Separate “accessibility” apps for people with disabilities always lag behind in features compared to apps that target a wider audience while still being accessible. Basically relegating people with disabilities to a second class experience.
289.
by thorum 3y ago
Public patents or not, can't imagine Apple is going to be thrilled with this guy for announcing to the world that their new product is secretly reading everyone's minds.
290.
by thorum 3y ago
hlky wrote most of the original code for the project, and might understandably think it’s “his” project. Setting aside whether his actions (which pissed everyone off and led to the code being rewritten from scratch without him under this ne
291.
by thorum 3y ago
Is there a concrete example of a back office job that can be replaced entirely by current generation LLMs, that couldn’t have already been replaced by a regular software program?
292.
by thorum 3y ago
I'd love to hear more about your approach for getting LLMs to understand how to use plugin commands. My own experiments have not worked very well (even vanilla ChatGPT through the gpt-3.5-turbo API doesn't seem to get the concept,
293.
by thorum 3y ago
Less about complexity than about how well-documented the subject is on the internet IMO. I’ve been using it to help me set up and troubleshoot AWS Elastic Kubernetes clusters, which are plenty complex, and I’d estimate it’s been around 95%
294.
by thorum 3y ago
From other discussions on Reddit, it seems that the upload function has been broken since March.
295.
Gfycat has been down for two days due to an expired SSL certificate(old.reddit.com)
96 points by thorum 3y ago|36 comments
296.
by thorum 3y ago
I’ve been playing since it came out here and there, probably at least 100-200 hours by now, and haven’t even defeated a single main boss. It probably tells you something positive about the game that one can get so much enjoyment just from r
297.
by thorum 3y ago
Someone in the replies asked GPT-4 to take a stab at it: https://twitter.com/ClintEhrlich/status/1647440753730420737
298.
by thorum 3y ago
I noticed this same issue with my AT&T internet. After some investigation, I found that using a VPN (Virtual Private Network) effectively resolved the problem. It seems that when internet traffic is encrypted and routed through a differ
299.
by thorum 3y ago
They’re probably just working toward a new image model powered by GPT-4’s image capabilities.
300.
by thorum 4y ago
While you make a reasonable point in terms of human psychology, when you create an AI, you are leaving the realm of human psychology and dealing with something new that doesn’t have to play by the same rules. I’d argue that’s a big differen

More ›