3 ms·
I do think LLM seems to work similar to what the left hemisphere of the brain does. The left hemisphere deals with an abstracted world broken into discrete elem
by theonemind 4y ago
I do think LLM seems to work similar to what the left hemisphere of the brain does. The left hemisphere deals with an abstracted world broken into discrete elements, and doesn't really make contact with the outside world--it deals with its system of representations. It also has a distinct tendency to generate bullshit, high suggestibility, and great respect for authority (which can apparently enter rules into its system of abstractions). The right hemisphere makes the contact with the outside world and does our reality checking, and it's really the more human element of us.
What this article says won't shock or disturb anyone deep into religious traditions with a strain of non-duality, which have had this message to shock and disturb people for thousands of years, in one way or another--there is no "you", especially not the voice in your head. I think you can come to a moment of intuitive recognition that the faculties of your brain that do reality checking aren't verbal, and they're riding shotgun to a bullshitter that never shuts up.
I think LLM can start looking more like automated general intelligence once it has some kind of link between its internal system of discrete abstractions and the external world (like visual recognition) and the ability to check and correct its abstract models by feedback from reality, and it needs an opponent process of reality-checking.
- rcarr 4y agoDo you have a blog at all? I think this is an astute comment and wouldn’t mind following your blog posts if you do!
- naijaboiler 4y agoThis whole left half, right half of the brain is very dodgy science. Yes there are functions that do have some sidedness, but that pop-sci right side/ left side dichotomy is mostly bunk
- theonemind 4y agoyou might find this worth checking out: https://www.amazon.com/Master-His-Emissary-Divided-Western/dp/0300245920 https://www.amazon.com/Master-His-Emissary-Divided-Western/d... he had to address this issue in the preface. The topic became a research career-ender after getting picked up by pop culture, but we do have solid science on hemispheric differences. The pop culture picture is, indeed, pretty wrong. It turns out that functions don't lateralize that strongly; they both tend to have the same capabilities, but operate differently.
- burnished 4y agoIt doesnt have any kind of internal representation of the world?
- lllllm 4y agoThe current systems like chatGPT actually have just such two parts. One is the raw LLM as you describe. The second one is another network acting as a filter on top of the first one. To be more precise, that second part is the process of finetuning with Reinforcement Learning from Human Feedback (RLHF). It trains a reward model to say if the first one was good or bad. Currently it's done very similarly to standard supervised learning (with human labelling) to say if the first model behaved good or bad, aligned or not with 'our' values. Anyway, while I remain sceptical about the roles of these in-flesh hemispheres, the artificial chatGPT-like systems indeed do have such left and right parts