One item posted to LinkedIn in April 2018 about an(other) unexpected result from research on AI, and another item posted in May on seven expected areas for AI advances in 2018. The potential of AI seems to be measured in the unexpected, while its value is calculated in the expected.
“AI Learns to Read Sentiment Without Being Trained to Do So,” Dom Galeon & Abby Norman, Futurism, 6 April 2017 (posted 17 April 2018)
Interesting research on an AI system: the training process for one task; the “emergent” property of the AI in the ability to distinguish positive & negative sentiment; & implications for learning systems generally.
The context of reading sentiment has me wondering about applications, & what might emerge from interaction between AIs with different aims (either complementary, such as in negotiating a deal, or competitive, such as resolving a conflict).
“Seven Artificial Intelligence Advances Expected This Year,” Forbes, 20 April 2018 (posted 19 May 2018)
Compilation by Forbes Technology Council of near term predictions of AI advances provides an interesting snapshot of priorities & directions. What paths are these leading us on? Anything missing from this list? Anything that should be but is not getting developed?
1. Improved Patient Health Outcomes (“Applying AI will utilize information from patients and providers to actively identify health conditions”)
2. Accelerated Improvement (“AI improves the more it’s used”)
3. Voice Assistants (“more intelligent and natural conversation between humans and voice assistants”)
4. Retail (“AI to enhance customers’ in-store experiences” & to “engage and interact with customers”)
5. Where It Makes Sense (“implement AI solutions where there are limited costs and clear returns”)
6. Online Chat (“a good way to engage buyers at their moment of interest”)
7. Communication (“voice recognition, dictation technology and communication” & “AI in combination with other technology such as virtual reality”)