AI
“ The magic of deep learning is you’ re adding more and more of these layers and you’ re trying to do more and more with the network”
Randall Hunt
CTO Caylent
Alexander, who created MIT’ s deep learning course back in 2015, points out how rapidly public benchmarks have shifted. Generative facial images were cracked between 2015 and 2018, followed by AI-generated video breakthroughs in 2020. Then came the public inflection point: ChatGPT( built on GPT-3.5) in late 2022, followed by GPT-4 in 2023.
“ The pace of this field is remarkable,” he says.“ It’ s easy for us to become desensitised to AI.”
The leap from GPT-3.5 to GPT-4 demonstrated to tech professionals and laypeople alike that the underlying capabilities of deep learning were scaling exponentially.
technologymagazine. com 153