5 ms·

An interesting idea for a website to track this stuff, especially because so much of the AI space is filled with vaporware and theoretical scenarios. One nitpi

An interesting idea for a website to track this stuff, especially because so much of the AI space is filled with vaporware and theoretical scenarios. One nitpick -- this incident is caused by traditional video game AI, not to be confused with the neural stuff that's causing waves right now: https://incidentdatabase.ai/cite/38/ https://incidentdatabase.ai/cite/38/

HlessClaudesman 3y ago
Yeah, this site needs a clear definition of what is an AI upfront. Good luck with that.
- inopinatus 3y ago
  Their minimum standard for inclusion appears to be any collection of if-statements.
  - pfsalter 3y ago
    This is the same standard for any Startup claiming to leverage AI, so I think that's fair
    - inopinatus 3y ago
      This measure of “fairness” merely impairs the utility of the data. Any standard based on puff-piece marketing is inherently garbage. There’s no research or policy value in dragging our thinking down to the same level. Unsurprisingly, many of their taxonomies are 30-50% labelled “Other”, a common warning flag of either junk data or lazy classifications.
      - lmcnulty 3y ago
        Note that the graphs on the taxonomy pages collapse labels outside the top nine under "all others." The incidents aren't literally labelled "other."
  - MacsHeadroom 3y ago
    It's even more loose than that! There are "incidents" of production lines not meeting quotas, requiring more manual labor. I guess conveyor belts are AI.
smcgregor 3y ago
(I am one of the people behind the website) We index everything that people believe is AI. Including GOFAI, for example: https://en.wikipedia.org/wiki/GOFAI https://en.wikipedia.org/wiki/GOFAI We also index incidents related to people transferring decision authority to simple programs like the Stanford vaccine distribution: https://incidentdatabase.ai/cite/91/ https://incidentdatabase.ai/cite/91/ In general, we err towards indexing useful information to the state of AI safety and introduced indexing "AI Issues" to develop the boundary and collect information on systems likely to produce incidents in the future: https://incidentdatabase.ai/apps/discover/?display=details&is_incident_report=false&page=1&sortBy=instant_search-en-featured https://incidentdatabase.ai/apps/discover/?display=details&i... Our editing guide is here: https://incidentdatabase.ai/editors-guide/ https://incidentdatabase.ai/editors-guide/

HlessClaudesman 3y ago

Yeah, this site needs a clear definition of what is an AI upfront. Good luck with that.

inopinatus 3y ago
Their minimum standard for inclusion appears to be any collection of if-statements.
- pfsalter 3y ago
  This is the same standard for any Startup claiming to leverage AI, so I think that's fair
  - inopinatus 3y ago
    This measure of “fairness” merely impairs the utility of the data. Any standard based on puff-piece marketing is inherently garbage. There’s no research or policy value in dragging our thinking down to the same level. Unsurprisingly, many of their taxonomies are 30-50% labelled “Other”, a common warning flag of either junk data or lazy classifications.
    - lmcnulty 3y ago
      Note that the graphs on the taxonomy pages collapse labels outside the top nine under "all others." The incidents aren't literally labelled "other."
- MacsHeadroom 3y ago
  It's even more loose than that! There are "incidents" of production lines not meeting quotas, requiring more manual labor. I guess conveyor belts are AI.

smcgregor 3y ago

(I am one of the people behind the website) We index everything that people believe is AI. Including GOFAI, for example: https://en.wikipedia.org/wiki/GOFAI https://en.wikipedia.org/wiki/GOFAI We also index incidents related to people transferring decision authority to simple programs like the Stanford vaccine distribution: https://incidentdatabase.ai/cite/91/ https://incidentdatabase.ai/cite/91/ In general, we err towards indexing useful information to the state of AI safety and introduced indexing "AI Issues" to develop the boundary and collect information on systems likely to produce incidents in the future: https://incidentdatabase.ai/apps/discover/?display=details&is_incident_report=false&page=1&sortBy=instant_search-en-featured https://incidentdatabase.ai/apps/discover/?display=details&i... Our editing guide is here: https://incidentdatabase.ai/editors-guide/ https://incidentdatabase.ai/editors-guide/