ETHAN MOLLICK on "the culmination of the first phase of the AI era ... [which] ends with the ... Google’s Gemini, the first LLM model likely to beat OpenAI’s GPT-4... enough pieces ... are in place ... to see what AI can actually do, at least in the short term... [although] implications of what this phase of AI will mean for work and education is currently unknowable" to everyone, including the teams building them, "but we can get a sense of its general shape".
At the acknowledged risk of anthropomorphizing:
- brains: we'll soon 1-2 other LLMs rival GPT-4, al;l capable of " out-innovate ...boost performance on complex tasks... [that] come with a lot of knowledge ... [but] have flaws ... problems with hallucinations... [so] not the dreaded/hoped-for Artificial General Intelligence"
- Nevertheless, even if all development stopped, it will probably still take us years to absorb the implications of what we have today
- vision give LLMs "a new method of interacting with the world... expanding their capabilities ... and uses ... [eg] read an operating manual to learn how to use a machine". This is however double-edged: combining facial recognition with recognising expressions, identifying locations and assessing contexts turns them into privacy invasion machines,
- voice: "gaining the ability to both listen and speak... can understand accents, mixes of languages... not bothered by crowded, noisy rooms". It also lowers psychological barriers between human and machine: "an oddly personal experience: ... feels like there is a real human interested in what you have to say".
- connect (LLMs to your own data): eg Bard connected to Gmail: "As AIs learn more about you, their usefulness will go up, though the full implications ...[ae] unclear".
What does this add up to? Already falling into place: AIs that can "talk to you, see you, know about you, do research for you, create images for you... personal assistant, intern, and companion ... [and potentially] troubling panopticon"