Rise of AI | ChatGPT, Dall-E, Midjourney etc

General Chat
322

I think there is a difference between causing no harm and outright refusing to discuss historical matters.

I fear we will agree to disagree here as I just don’t see any AI model that just won’t consider critical thinking around a government or historical incident to be dangerous and we should be cautious in singing praises over it.

Experts urge caution over use of Chinese AI DeepSeek | DeepSeek | The Guardian

We tried out DeepSeek. It worked well, until we asked it about Tiananmen Square and Taiwan | DeepSeek | The Guardian

I appreciate I may be in the minority these days but I do not trust the Chinese government to hold data securely or to prevent extreme censorship that goes beyond what reasonable people would consider acceptable.

1 Like
323

Well that would depend on whether discussion of that historical event could result in some kind of harm. I’m not a Chinese citizen so I don’t know enough to have an opinion, on what the consequences would be. I just know that ChatGPT similarly refuses to discuss prompts that might be controversial, sensitive, or frowned upon in our society.

Heck, our own access to the internet is pretty censored in its own right according to what our government deems we should and should not be able to access or learn about.

Don’t get me wrong, I’m very against censorship, generally, but I recognise the potency of the threat AI technology can pose when it isn’t reigned in, because long term, human misuse isn’t the problem, AI misusing AI is. But the censorship we’re discussing here is only being applied at the front end. The model quite happily begins reasoning and formulating a response before something trips and overrides the output. That means it isn’t baked into the model itself, so running it yourself, locally, won’t be subject to it. You’ll just need a bucket load of ram to run the full thing (retain the o1-like quality responses), or use an aggregate service (which right now seem to be outsourcing requests via API, so you might need to wait a bit).

Edit: just seen your edit -

I suspected your source might be The Guardian, and honestly, you couldn’t find a more biased one if you’d tried. To alleviate your China concerns further, the model does not phone home either. I’ve checked and verified that myself, running the smaller one locally on my Mac server. So you don’t need to trust China with your data if you use those models (not via the deepseek app or API), because it simply isn’t going to them.

The other benefit to the open source nature of the model, is that industry experts can independently vet it. So again, you don’t need to trust China. You just need to trust folks like me, and folks a heck of a lot smarter than me to assess it for you properly. That’s the one boon being FOSS (free and open source software) has over the proprietary models out of Silicon Valley, and I think it’s significant.

If it wasn’t FOSS, and the only access to the model was via the deepseek app (or API), then I’d be agreeing with everything you say in totality.

1 Like
324

I appreciate your detailed reply. I’d definitely be curious about any historical or government critiques that ChatGPT refuses to discuss.

I definitely do not think this is the same or in the same league, and it honestly concerns me how little people seem to care but at least I can tell you have considered your view on the matter.

For the avoidance of doubt: zero harm comes from knowing the truths about your country’s past; both good and bad. Nothing.

2 Likes
325

Hugging Face are on it!

1 Like
326

Yeah it just you, every western government, and everyone who knows anything about the CCP :smile:

3 Likes
327

I definitely have had too many chats where sympathy seems to be laying with China/Russia/Iran etc that I certainly feel in the minority.

Some folk have such a visceral hatred for western society and western governments that they take on the lies and mistruths of foreign governments almost as a contrary position as default.

The fact that I have to even debate some of the things this AI app won’t show is something I’d not think I’d have to do but here we are!

IMG_0847
IMG_08471206×1348 149 KB

And

IMG_0848
IMG_08481206×674 56.7 KB

And

IMG_0849
IMG_08491206×1780 214 KB

2 Likes
328

Regardless of open-source-AI, private-AI, territorial-AI, AI is here to stay and not on the terms of the big boys in the valley. Share prices will PLUMMET

The greatest threat to the US economy via tech in a long time. Wow.

1 Like
329

Can you guys even signup to Deepseek yet?

330

I think we need to establish a separation between the service provided by the creator of the model, and the open source model itself, because they’re ultimately two things really. And what’s exciting here is the model itself. That and the fact they’ve open sourced.

A British AI provider now can take this model, and launch an industry leading AI service hosted on British servers, without the pro-China censorship of the Chinese service which utilises the same model, thanks to its open source nature. They can also adapt it, improve it, contribute to it, fork it, etc. it’s FOSS. China need not be anywhere near it. They’re pretty irrelevant at this stage.

On a personal level, I really don’t care about the geopolitical debate surrounding these things. I’m pretty indifferent to all that crap. I’m just here for, and excited about, the innovation. Regardless of where it comes from. It’s the same thing with Electric cars. Outside of Tesla, there’s really only Chinese firms that are doing things half as exciting.

3 Likes
331

I have an account. But it’s no real difference (operationally) compared to the usual suspects:

Screenshot_20250128-222402
Screenshot_20250128-222402540×1202 212 KB

332

Ah nice, thought they paused it or something.

333

Agreed. This is the exciting part.

Hopefully the first of many open source models. Technology this impactful should be available to all, and not stuck behind a proprietary wall.

2 Likes
334

Eh? What happened to their own charter and principles? Do they only apply when they are winning the race?

OpenAI’s mission is to ensure that artificial general intelligence (AGI)—by which we mean highly autonomous systems that outperform humans at most economically valuable work—benefits all of humanity. We will attempt to directly build safe and beneficial AGI, but will also consider our mission fulfilled if our work aids others to achieve this outcome.

It’s also incredibly rich for them to start harping on about copyright at the same time they are asking various governments (including ours) to allow them to use copyrighted material to train their models.

Seems a threat to their share price means all principals go out the window.

2 Likes
335

I saw that this morning, very ironic from Open AI :joy:

1 Like
336

This is summary from the FT article

OpenAI claims Chinese AI start-up DeepSeek used its proprietary models to train its own open-source competitor, potentially breaching intellectual property rights. This practice, known as “distillation,” involves using outputs from larger models to train smaller ones, a common practice in the industry. OpenAI is investigating and taking countermeasures to protect its intellectual property, while also facing its own copyright infringement allegations.

337

The hypocrites!

The thing is, they’re likely banking on western governments taking a sympathetic and lenient stance favouring them for the simple reason the other guy is from China and we simply can’t have competition from china, can we?

This sort of response though is how threatened complacent incumbents have tried to snuff out a potential uprising in its crib instead of, you know, innovating. And they’ll lobby governments to that end. They’ll ultimately lose if they don’t win on innovation, as history shows us.

1 Like
338

Agreed, and also OpenAI is practically a baby, it’s way too early for them to act like a complacent incumbent!

1 Like
339

I’m curious about this (and lacking in knowledge). What is the benefit for this company in developing this model if its open source and basically anyone can then use the code to make their own model. How does this aid the company who made the model in making money?

Sorry if this is naïve question, but I’m genuinely curious as to why someone would bother if its not to licence it to others and make money off it.

340

There’s so many potential possible reasons why, and I don’t know their actual motive, so I’d just be speculating.

For what it’s worth, Apple and Meta both contribute heavily to open source AI efforts, so it’s not unheard of.

The Linux foundation has a good general answer to that question:

It is about organizations coming together to collectively solve common problems so they can separately innovate and differentiate

1 Like
341

Here’s Brandon!!

1 Like