4 ms·

Ask HN: Is our fear of AI just a fear of unreasonable people using AI?

3 points by acadapter 4y ago

uberman 4y ago
Bing has threatened someone and that is a GPT3+ "AI" that people kind of trusted. https://news.ycombinator.com/item?id=34804874 https://news.ycombinator.com/item?id=34804874 In my opinion it would be crazy to trust a self driving car, and AI or even a talking parrot with anything more than entertaining me without a human in the loop
- acadapter 4y ago
  The self driving car is, from my point of view, an example with people that use a low-quality control method with strange failure modes, and pretend like everything is okay. In many bad scenarios with AI, there's a person that granted AI some kind of decision power.
PaulHoule 4y ago
On one hand there is the "unreasonable" person who wants to get ChatGPT to write Hitler speeches or sadistic stories about sexual violence. On the other hand I'd say the Chinese Communist Party is much more "reasonable" and they'd like to use AI to support their surveillance state to stay in power. I am singling them out which isn't entirely fair because there are many actors out there who are less ethical and less responsible. For instance you have the tech support and romance scammers that could really benefit from ChatGPT's hypnotic ability. (That hypnotic ability is how ChatGPT impresses people with its writing skill despite frequently being completely wrong in terms of content.)
- quantified 4y ago
  On the spectrum is the automation of pseudo-human fake social media and journo media content for propaganda purposes that are pro-this-regime and anti-that-regime. Much more sophisticated troll farms, for example. I as a director in a state security apparatus, or someone for hire, can extend my reach very easily for or against some person or causes. It's not if it happens, it's when.
- sharemywin 4y ago
  To me the impressive part of GPT is that it gets anything right and it gets better just by making it bigger. The quality of the answers for DaVinci is way better because it stayed on topic. and got things right. yes it's not 100% accurate but there's no reason to assume it would have worked at all. why is chatgpt impressive? DaVinci ChatGPT is impressive because it can generate human-like conversations in real-time. It also offers natural language understanding capabilities and is capable of understanding and responding to complex questions. It is powered by the latest GPT-3 language model, which is the most advanced AI language model available today. curie Chatgpt is impressive because it is a very fast and easy to use messaging platform. It is also very user-friendly, making it a great choice for those who are new to messaging platforms. babage ChatGPT is impressive because it is one of the most robust chat applications on the market. It has a huge user base and is constantly being updated with the latest features. Ada There are a few reasons why chatgpt is impressive. First, it is a tool that is used daily by people in the business and government sector. This makes it a very important part of their workflows. They are easy to use and can be used in a number of different ways, which is also impressive. Additionally, chatgpt has been around for a long time and is still in use by a lot of the same teams. This makes it a very reliable tool that is likely to continue to work well in the future.
mindcrime 4y ago
I think most talk of "fear of AI" is a lot of overblown hype - at least the existential risk stuff. But to the extent that I see any reason for concern about AI, yes, it's more "fear of unreasonable people using AI". Aside from x-risk, my only concerns about AI involve the kinds of scenarios we've seen where AI systems merely continue to propagate existing human short-comings and flaws, like racial bias. Things like the facial recognition systems that don't work for Black individuals, or computer systems in the Justice system that treat people differently based on their ethnicity / gender / etc. Clearly there is still room for improvement in how we train and manage these systems so that they help correct our shitty biases and what-not, instead of perpetuating or amplifying them.
nh23423fefe 4y ago
This is a false dilemma. Are you afraid of nuclear fission or unreasonable people with nuclear fission. People state their fears on AI openly, you don't need to read into it. * AI is a tool of capitalism trained on stolen data to take jobs from humans It doesn't matter whether its plausible or true or coherent. It's a meme and it's spreading.

3 points by acadapter 4y ago

Bing has threatened someone and that is a GPT3+ "AI" that people kind of trusted. https://news.ycombinator.com/item?id=34804874 https://news.ycombinator.com/item?id=34804874 In my opinion it would be crazy to trust a self driving car, and AI or even a talking parrot with anything more than entertaining me without a human in the loop

acadapter 4y ago
The self driving car is, from my point of view, an example with people that use a low-quality control method with strange failure modes, and pretend like everything is okay. In many bad scenarios with AI, there's a person that granted AI some kind of decision power.

PaulHoule 4y ago

On one hand there is the "unreasonable" person who wants to get ChatGPT to write Hitler speeches or sadistic stories about sexual violence. On the other hand I'd say the Chinese Communist Party is much more "reasonable" and they'd like to use AI to support their surveillance state to stay in power. I am singling them out which isn't entirely fair because there are many actors out there who are less ethical and less responsible. For instance you have the tech support and romance scammers that could really benefit from ChatGPT's hypnotic ability. (That hypnotic ability is how ChatGPT impresses people with its writing skill despite frequently being completely wrong in terms of content.)

quantified 4y ago
On the spectrum is the automation of pseudo-human fake social media and journo media content for propaganda purposes that are pro-this-regime and anti-that-regime. Much more sophisticated troll farms, for example. I as a director in a state security apparatus, or someone for hire, can extend my reach very easily for or against some person or causes. It's not if it happens, it's when.
sharemywin 4y ago
To me the impressive part of GPT is that it gets anything right and it gets better just by making it bigger. The quality of the answers for DaVinci is way better because it stayed on topic. and got things right. yes it's not 100% accurate but there's no reason to assume it would have worked at all. why is chatgpt impressive? DaVinci ChatGPT is impressive because it can generate human-like conversations in real-time. It also offers natural language understanding capabilities and is capable of understanding and responding to complex questions. It is powered by the latest GPT-3 language model, which is the most advanced AI language model available today. curie Chatgpt is impressive because it is a very fast and easy to use messaging platform. It is also very user-friendly, making it a great choice for those who are new to messaging platforms. babage ChatGPT is impressive because it is one of the most robust chat applications on the market. It has a huge user base and is constantly being updated with the latest features. Ada There are a few reasons why chatgpt is impressive. First, it is a tool that is used daily by people in the business and government sector. This makes it a very important part of their workflows. They are easy to use and can be used in a number of different ways, which is also impressive. Additionally, chatgpt has been around for a long time and is still in use by a lot of the same teams. This makes it a very reliable tool that is likely to continue to work well in the future.

mindcrime 4y ago

I think most talk of "fear of AI" is a lot of overblown hype - at least the existential risk stuff. But to the extent that I see any reason for concern about AI, yes, it's more "fear of unreasonable people using AI". Aside from x-risk, my only concerns about AI involve the kinds of scenarios we've seen where AI systems merely continue to propagate existing human short-comings and flaws, like racial bias. Things like the facial recognition systems that don't work for Black individuals, or computer systems in the Justice system that treat people differently based on their ethnicity / gender / etc. Clearly there is still room for improvement in how we train and manage these systems so that they help correct our shitty biases and what-not, instead of perpetuating or amplifying them.

nh23423fefe 4y ago

This is a false dilemma. Are you afraid of nuclear fission or unreasonable people with nuclear fission. People state their fears on AI openly, you don't need to read into it. * AI is a tool of capitalism trained on stolen data to take jobs from humans It doesn't matter whether its plausible or true or coherent. It's a meme and it's spreading.