1/Several leaders share their hopes for AI in 2023, including finding key missing pieces that will enable algorithms to reason, building a personal data timeline, improving AI processes, discovering new principles for explainability, and using generative AI for active learning.
2/Yoshua Bengio wants to develop new architectures that can discover and reason with high-level concepts, rather than just brute force the learning process by scaling up existing models’ data and compute. deeplearning.ai/the-batch/yosh…
3/Alon Halevy envisions AI capturing massive amounts of data from your daily life -- photos, browsing, purchases -- and fusing it into a personal timeline that helps you track and achieve your goals, while preserving privacy. deeplearning.ai/the-batch/data…
4/Douwe Kiela points out key directions: Multimodality, grounding, and interaction so AI understands us better; alignment, attribution, and uncertainty to make models safer; data-centric AI to improve scaling; and better ways to evaluate AI models. deeplearning.ai/the-batch/douw…
5/Been Kim discusses AI explainability. AI has taken an engineering-centric approach, where researchers devise techniques via trial and error, and she urges developing fundamental scientific principles that make explanations more trustworthy and accurate. deeplearning.ai/the-batch/been…
6/Reza Zadeh sees generative AI bringing progress to active learning, where a system picks its own examples to be labeled to improve the data. With generative AI, he sees a potential revolution in algorithms generating new data to request to be labeled. deeplearning.ai/the-batch/reza…
1/Large language models like Galactica and ChatGPT can spout nonsense in a confident, authoritative tone. This overconfidence - which reflects the data they’re trained on - makes them more likely to mislead.
2/In contrast, real experts know when to sound confident, and when to let others know they’re at the boundaries of their knowledge. Experts know, and can describe, the boundaries of what they know.
3/Building large language models that can accurately decide when to be confident and when not to will reduce their risk of misinformation and build trust.