Our dystopia futures

Small changes can
affect millions of people instantly and change society forever

Some of the ethical issues inherent in AI

Autonomous Warfare; AI cyberattacks; bad actors; tools, research, data and models pooled into centralised organisations–monopolies; pervasive surveillance; job automation / job displacement; up skilling; using third party algorithms; company acquisition; team diversity for diverse outcomes; impacts on currently held freedoms e.g free speech, democracy; accountability; secrecy; when is it not appropriate to automate? unintended consequences; interpretable; explainable; transparent; impersonation; exploitation; false positives and true negatives; dogmatic systems; behaviour change / adjusting social norms; discrimination; failure states; model drift; bias data and models; appropriate data-set size; data reliability; data consent; personal data use; reproducing anonymised data; thresholds for acceptable model outcomes in any individual domain; Environmental impact of training models; Safety of new services; AGI goal alignment; AI Personhood

Autonomous Warfare; AI cyberattacks; bad actors; tools, research, data and models pooled into centralised organisations–monopolies; pervasive surveillance; job automation / job displacement; up skilling; using third party algorithms; company acquisition; team diversity for diverse outcomes; impacts on currently held freedoms e.g free speech, democracy; accountability; secrecy; when is it not appropriate to automate? unintended consequences; interpretable; explainable; transparent; impersonation; exploitation; false positives and true negatives; dogmatic systems; behaviour change / adjusting social norms; discrimination; failure states; model drift; bias data and models; appropriate data-set size; data reliability; data consent; personal data use; reproducing anonymised data; thresholds for acceptable model outcomes in any individual domain; Environmental impact of training models; Safety of new services; AGI goal alignment; AI Personhood

Autonomous Warfare; AI cyberattacks; bad actors; tools, research, data and models pooled into centralised organisations–monopolies; pervasive surveillance; job automation / job displacement; up skilling; using third party algorithms; company acquisition; team diversity for diverse outcomes; impacts on currently held freedoms e.g free speech, democracy; accountability; secrecy; when is it not appropriate to automate? unintended consequences; interpretable; explainable; transparent; impersonation; exploitation; false positives and true negatives; dogmatic systems; behaviour change / adjusting social norms; discrimination; failure states; model drift; bias data and models; appropriate data-set size; data reliability; data consent; personal data use; reproducing anonymised data; thresholds for acceptable model outcomes in any individual domain; Environmental impact of training models; Safety of new services; AGI goal alignment; AI Personhood

Autonomous Warfare; AI cyberattacks; bad actors; tools, research, data and models pooled into centralised organisations–monopolies; pervasive surveillance; job automation / job displacement; up skilling; using third party algorithms; company acquisition; team diversity for diverse outcomes; impacts on currently held freedoms e.g free speech, democracy; accountability; secrecy; when is it not appropriate to automate? unintended consequences; interpretable; explainable; transparent; impersonation; exploitation; false positives and true negatives; dogmatic systems; behaviour change / adjusting social norms; discrimination; failure states; model drift; bias data and models; appropriate data-set size; data reliability; data consent; personal data use; reproducing anonymised data; thresholds for acceptable model outcomes in any individual domain; Environmental impact of training models; Safety of new services; AGI goal alignment; AI Personhood

Autonomous Warfare; AI cyberattacks; bad actors; tools, research, data and models pooled into centralised organisations–monopolies; pervasive surveillance; job automation / job displacement; up skilling; using third party algorithms; company acquisition; team diversity for diverse outcomes; impacts on currently held freedoms e.g free speech, democracy; accountability; secrecy; when is it not appropriate to automate? unintended consequences; interpretable; explainable; transparent; impersonation; exploitation; false positives and true negatives; dogmatic systems; behaviour change / adjusting social norms; discrimination; failure states; model drift; bias data and models; appropriate data-set size; data reliability; data consent; personal data use; reproducing anonymised data; thresholds for acceptable model outcomes in any individual domain; Environmental impact of training models; Safety of new services; AGI goal alignment; AI Personhood

Autonomous Warfare; AI cyberattacks; bad actors; tools, research, data and models pooled into centralised organisations–monopolies; pervasive surveillance; job automation / job displacement; up skilling; using third party algorithms; company acquisition; team diversity for diverse outcomes; impacts on currently held freedoms e.g free speech, democracy; accountability; secrecy; when is it not appropriate to automate? unintended consequences; interpretable; explainable; transparent; impersonation; exploitation; false positives and true negatives; dogmatic systems; behaviour change / adjusting social norms; discrimination; failure states; model drift; bias data and models; appropriate data-set size; data reliability; data consent; personal data use; reproducing anonymised data; thresholds for acceptable model outcomes in any individual domain; Environmental impact of training models; Safety of new services; AGI goal alignment; AI Personhood

Autonomous Warfare; AI cyberattacks; bad actors; tools, research, data and models pooled into centralised organisations–monopolies; pervasive surveillance; job automation / job displacement; up skilling; using third party algorithms; company acquisition; team diversity for diverse outcomes; impacts on currently held freedoms e.g free speech, democracy; accountability; secrecy; when is it not appropriate to automate? unintended consequences; interpretable; explainable; transparent; impersonation; exploitation; false positives and true negatives; dogmatic systems; behaviour change / adjusting social norms; discrimination; failure states; model drift; bias data and models; appropriate data-set size; data reliability; data consent; personal data use; reproducing anonymised data; thresholds for acceptable model outcomes in any individual domain; Environmental impact of training models; Safety of new services; AGI goal alignment; AI Personhood

Autonomous Warfare; AI cyberattacks; bad actors; tools, research, data and models pooled into centralised organisations–monopolies; pervasive surveillance; job automation / job displacement; up skilling; using third party algorithms; company acquisition; team diversity for diverse outcomes; impacts on currently held freedoms e.g free speech, democracy; accountability; secrecy; when is it not appropriate to automate? unintended consequences; interpretable; explainable; transparent; impersonation; exploitation; false positives and true negatives; dogmatic systems; behaviour change / adjusting social norms; discrimination; failure states; model drift; bias data and models; appropriate data-set size; data reliability; data consent; personal data use; reproducing anonymised data; thresholds for acceptable model outcomes in any individual domain; Environmental impact of training models; Safety of new services; AGI goal alignment; AI Personhood

I'm just
a developer,
a designer,

a business

We're all part of the solution

“The best way to predict the future is to invent it.”

Alan Kay

Thank you

www.machine-ethics.net

www.BenByford.com

@BenByford