Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by reissbaker on Monday January 27, 2025 @12:32PM
(#42843530)
It's a 7B "unified model" LLM/VLM (not a diffusion model!) that out-benchmarks Dall-E 3 and Stable Diffusion Medium. It's released under the DeepSeek License, which is pretty-open license that allows commercial use but restricts military use, along with a few other content-based restrictions.
[ Parent
| on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by lenerdenator on Monday January 27, 2025 @12:34PM
(#42843546)
> restricts military use
I'm sure the powers-that-be will absolutely pay attention to that clause.
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by operator-name on Monday January 27, 2025 @02:50PM
(#42844906)
You could say the same for the GPL, yet it's wording is enough to curb adoption from corporations.
Large organisations like the military have enough checks and balances to avoid these kind of licences with a 10ft pole.
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by qwertox on Monday January 27, 2025 @01:17PM
(#42843998)
Yeah, they should! Not that the missile then makes a 180° turn to "return to sender" because it noticed that the target is a Chinese military base.
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by culi on Monday January 27, 2025 @01:28PM
(#42844094)
The code is open sourced
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by jprete on Monday January 27, 2025 @01:32PM
(#42844137)
There's no meaningful inspection of LLM code, because the real code is the model weights.
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by mschoening on Monday January 27, 2025 @01:32PM
(#42844135)
See Sleeper Agents (https://arxiv.org/abs/2401.05566).
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by cosmojg on Monday January 27, 2025 @03:17PM
(#42845157)
Who in their right mind is going to blindly take the code output by a large language model and toss it on a cruise missile? Sleeper agents are trivially circumvented by even a modicum of human oversight.
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by carimura on Monday January 27, 2025 @01:32PM
(#42844131)
but what about training data?
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by culi on Tuesday January 28, 2025 @02:23AM
(#42849756)
The weights and data pipeline are open sourced and described explicitly in the paper they published. The non-reasoning data isn't nearly as interesting as the reasoning data though
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by Aaronstotle on Monday January 27, 2025 @12:36PM
(#42843562)
How are these licenses enforceable?
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by reissbaker on Monday January 27, 2025 @12:39PM
(#42843602)
Lawsuits, but it's mainly just CYA for DeepSeek; I doubt they truly are going to attempt to enforce much. I only mentioned it because it's technically not FOSS due to the content restrictions (but it's one of the most-open licenses in the industry; i.e. more open than Llama licenses, which restrict Meta's largest competitors from using Llama at all).
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf]
by jiggawatts on Monday January 27, 2025 @07:51PM
(#42847659)
I've always wondered why nobody has tried to scale image-generation models to modern LLM sizes, such as 200-500B parameters instead of 1-7B...