Niantic, the company behind the extremely popular augmented reality mobile games Pokémon Go and Ingress, announced that it is using data collected by its millions of players to create an AI model that can navigate the physical world.
In a blog post published last week, first spotted by Garbage Day, Niantic says it is building a “Large Geospatial Model.” This name, the company explains, is a direct reference to Large Language Models (LLMs) Like OpenAI’s GPT, which are trained on vast quantities of text scraped from the internet in order to process and produce natural language. Niantic explains that a Large Geospatial Model, or LGM, aims to do the same for the physical world, a technology it says “will enable computers not only to perceive and understand physical spaces, but also to interact with them in new ways, forming a critical component of AR glasses and fields beyond, including robotics, content creation and autonomous systems. As we move from phones to wearable technology linked to the real world, spatial intelligence will become the world’s future operating system.”
By training an AI model on millions of geolocated images from around the world, the model will be able to predict its immediate environment in the same way an LLM is able to produce coherent and convincing sentences by statistically determining what word is likely to follow another.
I’ll copypaste an interesting comment here:
You don’t even need robots replacing humans, or people believing so. All you need is people feeling that you’re profiting at their expense.
Also obligatory “If you’re not paying for the product, then you are the product”.
I’ve found myself thinking “well, you just helped teach the AI about that one…” various times when reading content online.
It’s a strange thing to know that a form of the basilisk is real. Things posted will help AI get better, if only my teeny tiny increments each time.
AI learning isn’t the issue, its not something we will be able to put a lid on either way. Either it destroys or saves the world. It doesn’t need to learn much to do so besides evolving actual self-agency and sovereign thought.
What is a huge issue is the secretive non-consentual mining of peoples identity and expressions.
And then acting all normal about It.
So… there is no Artificial Intelligence. The AI cannot hurt you. It is just a (buggy) statistical language parsing system. It does not think, it does not plan, it does not have goals, it does not understand, and it doesn’t even really “learn” in a meaningful sense.
If we’re talking about machine learning systems based on multi-dimensionl statistical analyses, then it will do neither. Both extremes are sensationalism and arguments based on the idea that either such outcome will come from the current boom of ML technology is utter nonsense designed to drive engagement.
Oh, is that all?
No one on the planet has any idea how to replicate the functionality of consciousness. Sam Altman would very much like you to believe that his company is close to achieving this so that VCs will see the public interest and throw more money at him. Sam Altman is a snake oil salesman.
This is absolutely true and correct and the collection and aggregation of data on human behavior should be scaring the shit out of everyone. The potential for authoritarian abuses of such data collection and tracking is disturbing.
I didn’t say it was an issue. I just said it was a strange feeling to know AI is watching us talk past each other.
I sort of misread your comment as saying the basilisk is inevitable which is a thought i would describe as least oopsie-issue-level.
Still there are many other people bent on directly poisoning AI to counteract the learning but i just fear that will get it to dangerously rogue mentally challenged AI faster then if we aimed for maximum coherent intelligence and hope that benevolence is an emergent behavior from it.
But more at hand. If we build AI by grossly exploiting our own fellow-humans. How do we expect it will treat us once it reaches a state of independent learning.
The only down side, IMO, is that the models are proprietary and closed.
What are platforms going to do about it? Start to demonetize AI generated videos and ban AI written fan fiction?
I think people will still “contribute” because they also don’t care that their use of certain platforms leaks data used to target ads at them.
In the same vein though, once AI essentially destroys a site like Stack Overflow, where will AI companies source new training data with updated information? Also, we are likely to see something like 50% of content being AI generated. Are AI models then going to train on the content they themselves created? What is the impact of that? What is the use?
It leads to model collapse. The second AI starts to focuses on certain patterns in the output of the first AI instead of the actual content and you get degraded output. They are pattern matching machines after all. Repeat the cycle a few times and all output becomes gibberish. Think of it as data incest.
So the AI companies are pretty desperate for more fresh user data. More data is the only way they have currently to push through the diminishing returns.