Language Is Not All You Need: Aligning Perception with Language Models
1–10 of 29 posts
Re: Language Is Not All You Need: Aligning Perception with Language Models
#2Can someone ELI5 me why suddenly LLM and transformers became all the rage in AI scene?
Re: Language Is Not All You Need: Aligning Perception with Language Models
#3Can someone ELI5 me why suddenly LLM and transformers became all the rage in AI scene?
I believe it's because they became more capable
Re: Language Is Not All You Need: Aligning Perception with Language Models
#4Can someone ELI5 me why suddenly LLM and transformers became all the rage in AI scene?
Transfomers are currently state of art i.e. the problems that transformers are currently solving cant be solved better by any other known technique/algorithm. Thats the main reason why they are so popular at the moment. LLMs are popular right now because a recent transformer based neural network has proven to be fun/useful i.e. gpt-3/chatgpt. A lot more useful than previous language models at least.
Re: Language Is Not All You Need: Aligning Perception with Language Models
#5First image that popped to mind were the "squids" from Matrix.
Re: Language Is Not All You Need: Aligning Perception with Language Models
#6Can someone ELI5 me why suddenly LLM and transformers became all the rage in AI scene?
Transformers were used on text to train models without needing labeled data. People realized that simply scaling the data and models meant better performance. when they scaled even further, emergent intelligence started appearing and the models were dominating every known task. Now everyone wants an LLM
Re: Language Is Not All You Need: Aligning Perception with Language Models
#7Multimodal transformers is the logical leap forward. These ideas are not new, deep mind and google have been at this for some time. It’s yet another arms race and those with the most diverse data will win
Re: Language Is Not All You Need: Aligning Perception with Language Models
#8The chain-of-thought prompting in section 4.5 is extremely interesting to me, but it looks like they're missing a test group - what is the performance if the image is simply described and then the task is evaluated using only the text of the description, not only when combined with the image.
Re: Language Is Not All You Need: Aligning Perception with Language Models
#9Can someone ELI5 me why suddenly LLM and transformers became all the rage in AI scene?
Transfomers are currently state of art i.e. the problems that transformers are currently solving cant be solved better by any other known technique/algorithm. Thats the main reason why they are so popular at the moment. LLMs are popular right now because a recent transformer based neural network has proven to be fun/useful i.e. gpt-3/chatgpt. A lot more useful than previous language models at least.
RWKV is showing that maybe RNNs can perform on par with transformers
Re: Language Is Not All You Need: Aligning Perception with Language Models
#10Finally the Metaverse is taking shape. Those human friends on Facebook can now be replaced with Metafriends