Slashdot Mirror
Tech Companies Pledge To Use Artificial Intelligence Responsibly (axios.com)
An anonymous reader shares a report: The Information Technology Industry Council -- a DC-based group representing the likes of IBM, Microsoft, Google, Amazon, Facebook and Apple -- is today releasing principles for developing ethical artificial intelligence systems. Why it matters: The tech industry is trying to get ahead of growing anxieties about the societal impact of AI technologies, and this is an acknowledgement on companies' part that their data-hungry products are causing sweeping changes in the way we work and live. The companies hope that pledging to handle this power responsibly will win points with critics in Washington, and that showing they can police themselves will help stave off government regulation on this front. The principles include: Ensure the responsible design and deployment of AI systems, including taking "steps to avoid the reasonably predictable misuse of this technology by committing to ethics by design." Promote the responsible use of data and test for potentially harmful bias in the deployment of AI systems. Commit to mitigating bias, inequity and other potential harms in automated decision-making systems. Commit to finding a "reasonable accountability framework" to address concerns about liability issues created when autonomous decision-making replaces decisions made by humans.
The most serious dangers from AI come from the inability to predict or control it. I'm not concerned someone is going to create an AI to wipe out humanity, I'm concerned about side effects from complex optimization algorithms that are doing exactly what we ask them to do.
Using an AI to adjust tax policies to reduce hunger might not reduce it in the way people desire.
OK, it's safe to assume that they'll take some precaution in building your AI toaster. Your home thermostat. Your smart vibrator... There isn't much financial incentive to do evil.
But wouldn't it be tempting to bid on a 5 billion dollar contract for weaponized AI ? Every government will want one.
...omphaloskepsis often...