harrisoned

I noticed that same behavior across older Gemini models. I build a chatbot at work around 1.5 Flash, and one day suddenly it was behaving like that. it was perfect before, but after it always saluted the user like it was their first chat, despite me sending the history. And i didn't found any changelog regarding that at the time.

After that i moved to OpenAI, Gemini models just seem unreliable on that regard.

I agree its an amazing payment method, it worked for me for most of the time. Still, we depend on bank's stability and technical availability for it to work. Once i needed to pay for something and forgot my card at home, at that same time my bank was going trough technical issues and i couldn't pay.

Despite rare reliability issues, my fear about it is that it requires a phone. Being so popular, i fear when places will refuse any other form of payments and accept only PIX, making anybody not using a phone unable to buy their products, with the common assumption that everybody uses it ("don't you guys have phones???"). You can't install banking apps on rooted phones or alternative mobile OSs (or is very very hard), so you are trapped with Android or IOs to use it.

I hope it doesn't come to that, but it seems it's going that way.

It is. A friend of mine has this. It took her so much time to actually seek help, despite the excruciating pain, and she got part of her digestive system removed because it spread too much.

Nowadays, after 2 years of the surgery, she manages it with a restrict and healthy diet, and the pill. It takes a big toll on the well being even after the pain is gone, and she is almost always tired because of it, the body is constantly fighting the inflammation.

I have a big issue with this, and the truth is that the majority of people simply do not care and/or do not understand the implications.

By tying your service to a smartphone your are basically refusing to provide service if the costumer doesn't agree to Apple's or Google's TOS. If the app doesn't complain about emulation or something different than Android or IOS you are in luck, but that's not the case with most banking apps. And that's only talking about people who don't have it by choice and have money to buy one.

For me, once, it went beyond: I took my first dose of the Covid vaccine, and the second dose's date would still be announced. I asked where it would be available to the nurse, "On the Instagram page of the <local health body>". "But i don't have Instagram" i said, and the nurse shrugged. It requires both a phone and a social media account with your real info, but since absolute nobody complains about it they just do because it's easier.

This will continue as long people are complacent with it. In some places the government is required to provide you services, by law, by any means available and not depending on 3rd party service, but they do require apps anyway and people stay quiet. Phones as an alternative is fine, it's a tool, but should not be an obligatory device for you to be considered an human being.

I agree with that. At work, we are about to implement a decent LLM and ditch Dialogflow for our chatbot. But not to talk directly to the client (it's asking for a disaster), just to recognize intentions, pretty much like Dialogflow but better.

Right now there are many small but decent models available for free, and cheap to use. If it wasn't for the hype, it would never have reached that level of optimization. Now we can make decent home assistants, text parsers and a bunch of other stuff you already mentioned.

But someone paid for that. The companies who believed this would be revolutionary will eventually have a really hard reality check. Not that they won't try and use it for critical stuff, but once they do and it fails spectacularly they will realize a lot of money went down the drain.

It certainly has use cases, just not as many as the hype lead people to believe. For me:

-Regex expressions: ChatGPT is the best multi-million regex parser to date.

-Grammar and semantic check: It's a very good revision tool, helped me a lot of times, specially when writing in non-native languages.

-Artwork inspiration: Not only for visual inspiration, in the case of image generators, but descriptive as well. The verbosity of some LLMs can help describe things in more detail than a person would.

-General coding: While your mileage may vary on that one, it has helped me a lot at work building stuff on languages i'm not very familiar with. Just snippets, nothing big.

I'm curious about the ToS. As far as i know, their version of ChatGPT on Copilot don't like +18/sexual stuff, so how would the Copilot+ react to a lot of it? Would you get banned? would it work? Would it simple ignore all the stuff when asked for it?

I did that with Llama 3 8B with some stuff i could think of, and it did very good. It was on par with GPT4. I prompted some scenarios and asked it to use CoT. Scenarios like "i was standing and eating chocolate, and it melted. Will i find chocolate at my feet?", and the reasoning was pretty good.

But there was something it did way better than GPT4. I asked to create 10 phrases where the last word was an animal, excluding equines, and in alphabetical order. GPT3.5 and GPT4 aren't able to follow such instructions, but the 8b model did it with maestry.

I personally don't use LLMs to code, besides a few snippets or when i'm out of ideas what/how do to. But models like Starcoder and Code Llama is what i see people often using for this purpose. There are benchmarks for various languages, you can find those on Hugging Face.

If your intention is coding something complex, you should try a model finetuned for that, i don't think llama 3 is that good for coding. But i used standart prompt engineering stuff, same as from llama 2. Instead of chat, you use a completion mode, where you just need to give it a text to continue writing from.

I'm playing with the llama 3 8b instruct model out of curiosity, and it is insanely better than the llama 2 on that regard. It's almost like a fully uncensored model. it did refused to make pentest scripts when i asked, which is fine. But it made the scripts after i changed the system prompt to something more 'permissive'. The model seem to adhere more to user commands, and it's more useful overall. It's even good at complex math, which is insane considering even GPT4 is bad at it.

I wasn't sure if meta would release the model to the public, i'm glad they did.

Just like many here have said already, GPT4 is being useful for coding for me. It is an amazing parser, specially, and save me precious time. Of course it's not able to do anything on it's own or without supervision, but is has been better than looking up to examples on Google.

I also have been experimenting with it to replace the intention classifier part of Google's dialogflow. We use it at work for our chatbot. Earlier, we used Watson and it was amazing, but became very expensive. Dialogflow is cheap, but it is as innacurate with complex natural language as it is cheap.

Mixtral (8x7B) has proved extremely accurate in identifying intentions with a consistent JSON output, giving it a short context, so i assume a simple 7B model would do the job. I still don't know if it is financially worth it, but it's something i'm gonna try if i can't fix the dialogflow's intentions. But in no way the model's output would directly interface with a client. That's asking for trouble.

I wonder if they where really offended or it was an impulsive overreaction on their part :P

In times where people like that exists, chatbots that can swear are weapons of mass destruction.

> There is a good chance they might sue OpenAI for it.

Indeed there is! Companies have been sued (and lost) for way less than that. But it begs the question: should the user not know better? Is it really the company's fault, even with gigantic disclaimers in the front page and the user clicking in "i read it and i agree"?

For me, this is the issue. Doesn't matter if there is a warning, an advice or even a manual, the company will ALWAYS be the one to blame for anything. Treating users like that has made way for people to act in bad faith so it makes sense for them to act like that. Better safe than sorry.

I get it that companies want to distance themselves for bad stuff those LLMs output. But what baffles me is the way they deal with this issue.

Nowadays if you share or reproduce anything controversial, is as if you are endorsing such content. LLM's outputs are user generated content for what is worth it, both from inferece and from training. That line of thinking that the host would be responsible for everything it says, even if the users force it to, is ludicrous to me but it seems like how the world nowadays has become.

Take the headlines from when BingAI was acting-up, the level of anthromorphization of those applications were insane. I suppose they do it for the headlines, but people also falls on that and play along. If ChatGPT says something bad, is as if the CEO of OpenAI was saying it verbatin, period, and it blows back directly at the company. Take a look at the subreddit of ChatGPT and see how people act when it outputs something weird.

It saddens me that people lack common sense in such a way, and that's why we can't have nice things in the end. The companies also assumed the roles of babysitting the users, like a bunch of no-brainers. As long as online culture follow this, it will get worse.

I think that is the biggest reason for the excessive safety as well. Just like all services try to be safe for those same things.

What do puzzles me is why those companies act that way. Is it because costumers are dumb enough to not differentiate "user generated content" from "a company endorsing such content", or their shareholders actively want to push such moral values?

Well, "safely" is subjective i guess. This means an AI that will not "get rogue and kill us all" or an AI that will not swear back at the user or generate unethical content even when prompted?

I am really not concerned about the first aspect. The second one is a lot harder to guarantee considering how those LLMs work. People will push them to the limit, and when there is no one to blame, they will blame the company who made to product. I am concerned that they will sacrifice usability, accuracy, and basically neuter the product because of that notion of safety. It's already happening, especially with Dall-e 3 where they pre-process your prompts at the API endpoint, that shows how scared they are about it being misused and how bad it can be to them, as the user can't be responsible for their own prompt. Bulding a safe complex tool like that, in that sense, that is devoid of any means of misuse is very very hard to do without making it bleak. I really hope something changes along the way to fix that.

It is. But wasn't worth the trouble to fight over it.

Here, in Brazil, this is becoming more and more common, but in the US it's already an endemic problem for those who don't use tech.

My biggest issue with all this comes to the forced dependency of 3rd party services and ToS agreements to use something totally unrelated to it, along with all the technical issues that, we all know, happens every time. For example, here the federal bank will ask you to install their app trough App/Play store. Since there are still MANY tech illiterate people here, they still have in-person service. But for the majority of people, they will try to force you to use the app. It's a "don't you guys have phones?" issue i see spreading everywhere, but almost nobody complains about it because... it's easier. They don't care about privacy, data collection, none of this, and companies (with government bodies) are taking advantage of this for reasons i rather not speculate about here.

During Covid, i took my first dose of vaccine and the nurse said "check out our Instagram to see the next date of vaccination". "But i don't have a Instagram account or any social media. Is there any official channel, like a .gov website?" i asked, and she shrugged like 'its your problem then'. I walk out of any restaurant that forces you to scan a QR code for the menu, and if possible, i avoid the usage of my phone all together. As long as people are compliant in that behavior, things will only get worse.

Nine years ago, during the interview with the manager of a textile company, he asked what my religion was. I said i didn't had any, and then he said "You know what happens with people who don't follow any religion, right?" and pointed down, indicating that i was 'going to hell'. That made me want to quit right there, but i was so much in need of a job.

I wasn't hired in the end, despite doing well on the tests. But it was for the best as everything worked out in the end.

I started watching Farscape a few weeks ago after seeing many recommendations, critics, and because it's free to watch on Plex. Now i see why people say what they say.

I think Meta did a very good job with Llama2, i was skeptical at first with all that talk about 'safe AI'. Their Llama-2 base model is not censored in any way, and it's not fine-tuned as well. It's the pure raw base model, i did some tests as soon as it released and i was surprised with how far i could go (i actually didn't get any warning whatsoever with any of my prompts). The Llama-2-chat model is fine-tuned for chat and censored.

The fact that they provided us the raw model, so we could fine-tune on our own without the hassle of trying to 'uncensor' a botched model, is a really great example on how it should be done: give the user choices! Instead, you just have to fine-tune it for chat and other purposes.

The Llama-2-chat fine-tune is very censored, none of my jailbreaks worked, except for this one[1], and it is a great option for production.

The overall quality of the models (i tested the 7b version) has improved a lot, and for the ones interested, it can role-play better than any model i have seen out there with no fine-tune.

1: https://github.com/llm-attacks/llm-attacks/

Non-tech people can use it with no issues, my whole company uses it. People just need to pay attention to what they are doing, and be organized. At the beginning it can be kinda painfull, but it's all worth it.

I work at a small-to-medium sized ISP, and we use pure Redmine since 2018 for documentation, and 'internal tickets'. It's a really good tool, helped us get really organized. You can customize the heck out of it, nowadays we have really complex BIs to monitor tickt times, SLAs, project activity, etc. The API is very good and i made a ton of usefull things with it, even Telegram bots for our internal groups.

The biggest complaint we have on it is how bad it's wiki search features are, and indeed they are too simple.

While i think bots and AI is indeed widespread, i think this percentage is accurate only in very specific places. The internet isn't dead, sometimes we (tech people) don't take in consideration how much of the world actually uses the web for things/in ways we wouldn't even consider (bad and good), and how much of it contributes nothing and easily passes as just background noise for us and could, easily, passes as a bot writing.

i don't know if this is real in France, but here in Brazil, there is a bill trying to do the same thing (hasn't passed yet). Social media would be obligated to pay news outlets, and on the Chapter VII, Article 32, § 6º it says:

> "The provider may not promote the removal of journalistic content made available in order to exempt themselves from the obligation referred to in this article, except for the cases provided for in this Law, or by court order specific."

This is what i mean by underlying problems. Every extremist organization lures people with manipulative tactics, and if you are not in a good state of mind to spot those, you will be caught. That's why education and knowledge about those subjects is important, so people can understand and think about them. If you never understood why they are so bad, you would rely on others to never see it on your screen.

> Why is the blame on people for what is made available to them? This wouldn't be 'protecting people from their own mental and moral malleability,' this would be preventing others from making this which abuse those malleability for profit.

I do believe our own actions are the only thing we have absolute control in life. In a perfect world, maybe such efforts would be more effective. I take in account the subjectivity of human actions and behavior in that case. The amount of safeguards you would need to put in order to protect people in that scenario would leak into many other areas, or create an eternal conflict at minimum (it isn't like we don't have one already...). I believe that education and knowledge is the path to solve such issues, because it creates a mechanism for people to deal with them (like a vaccine), not to never see them.

> The same type of argument you have made here could be made for legalizing ponzi schemes. People _should_ know that there is no quick rich scheme and the history of these schemes are full of fraudsters. Is it wrong to ban this type of abuse because their victims made the choice to engage in the first place?

People should be educated about it. It wouldn't solve all the problems immediately, but it is more efficient if people protect themselves from these issues. Ponzi schemes are illegal and they still an issue, the same would happen to other regulations in my opinion. Companies always find a way to profit.

Social media is a moneymaking machine, and data is valuable. It's not just internet fun, it's serious business. In a way, i don't think they are too far apart. I dislike social media and haven't used it in many years. You could apply regulations, is not that it is wrong, it's just ineffective against underlying issues and the problem itself in the long term. The effort should be more towards a long term solution. The line here is blury, that's what happen when you deal with humans. You could go towards the "protect people from harm" or "give them tools for self-protection" idea, i honestly prefer the latter.

> Do they?

They do. Can they be influenced? Yes. Can they be coerced trough social pressure? Yes. Can they fall for targeted advertisement and/or manipulation? Yes as well. Does this exclude the responsibility of their own actions? I don't think so. We can argue that this problem was born because of their parents or their own lack of knowledge/critical thinking later in life.

> Where are you from?

I am brazilian.

> The hard fact that a community where our 'median commenter' is above the median in the general population hates to realize is that roughly 1/3 or so of all people are just not remotely the type of person who can do the abstract meta-cognition to see they're being manipulated.

True! This is a complex subject that i have been interested in a long time. It all boils down to 'digital inclusion' (making services dumber/simpler to be easier for the average person) for profit, and it's long term effects. But that's another whole can of worms.

There is no easy solution for that problem.