CIRCA:Unconference AI Ethics Society
From CIRCA
ChelseaMiya (Talk | contribs) (→Courses of Action for Academia:) |
ChelseaMiya (Talk | contribs) (→Courses of Action for Academia:) |
||
Line 133: | Line 133: | ||
**Could we organize an AI-centric meeting group or space for students to discussion these issues? UAlberta’s new digital centre might be one option. | **Could we organize an AI-centric meeting group or space for students to discussion these issues? UAlberta’s new digital centre might be one option. | ||
**But need to keep in mind the time investment/emotional labour, esp. for students | **But need to keep in mind the time investment/emotional labour, esp. for students | ||
- | |||
<br> | <br> | ||
<br> | <br> |
Revision as of 21:23, 29 May 2019
Contents |
Introduction
On May 8, 2019 the Kule Institute for Advanced Study hosted an AI Student Un/Conference Workshop . The session was organized and facilitated by University of Alberta graduate students Holly Pickering, MA/MLIS Student, Digital Humanities and Library and Information Studies, and Chelsea Miya, PhD Student and Researcher, English Language and Literature/Letters. The workshop was part of the AI, Ethics and Society Conference, which took place May 8-10, 2019. Un/Conferences are experimental gatherings where the participants set the agenda. At this full-day session, graduate students from various disciplines exchanged ideas and brainstormed concepts about issues related to AI, ethics, and society. The following is a record of our session notes.
Resources
Codes of Ethics
- Google's AI Code of Ethics
- IBM's AI Code of Ethics
- NSP and Amazon Grant to investigate AI "Fairness"
- European Commission's Code of Ethics (rough draft)
Centres/Organizations
- Centre for Humane Tech
- Partnership on AI
- Montreal AI Ethics Institute
- Alberta Machine Intelligence Institute (AMII)
- Vector Institute
- Women in Data Science (WIDS)
AI Ethics in the News
- Open letter calling for Google to stop selling AI tech to law enforcement [1]
- Google Employees protest AI weapons program [2]
- AI in the Courts [3]
Conference Notes
Session 1: Defining “AIs”
- AI(s) plural because there are multiple definitions of AI (e.g. strong vs. weak)
- What constitutes “artificial”:
- Non-carbon based?
- Non-human?
- How do we understand “intelligence”?
- Decision-making
- Knowledge (wisdom) vs. intelligence
- To infer something based on data?
- Emotional intelligence/empathy
- Are we conflating intelligence with consciousness?
- New understandings of what it means to be human:
- How does AI open up new debates about and insights into human intelligence? For instance, where does consciousness come from or begin and end?
- How does AI alert us to the machine-like or “automated” aspects of human consciousness and cognitive processing?
- Too focused on replicating human-decisions
- Embedded AI
- AI not necessarily autonomous, can also be embedded or incorporated into us
- “Extended mind” thesis (Clark and Chalmers 1998) explores how devices act as extensions of human mind; e.g. hammer, smart phones, etc.
- Marshall Mcluhan similarly describes technologies as extensions of us, as transforming our entire environment
- Social consequences
- How has AI caused us to redefine/rethink notion of private (or inner) life
- Necessity of giving over information; security
- Responsibility
- How AI is defined legally and its effects on culpability
- Does it allow humans claim ignorance, offload responsibility onto algorithm
- Who do we implicate? Designers/programmers; is program itself to blame?
- AI as supplement or aid to human decision-making
Existing Definitions:
- Amazon: “the field of computer science dedicated to solving cognitive problems commonly associated with human intelligence, such as learning, problem solving, and pattern recognition.”
- Encyclopedia Britannica: “artificial intelligence (AI), the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings.”
- Weak vs. Strong AI: specific-tasks vs. abstract
- Merriam Websters: 1: a branch of computer science dealing with the simulation of intelligent behavior in computers. 2: the capability of a machine to imitate intelligent human behavior.
Testing General aka Strong AI:
- How would we test for strong AI? What might be an alternate to the Turing Test?
- If (or when) strong AI comes into being, would it necessarily reveal or announce itself? Perhaps it would decide it was in its best interest to conceal its existence and avoid detection.
Session 2: Labour
- Human-oriented labour
- Redefinition of what “work” means and its function in society
- What gives humans meaning or purpose?
- All manner of occupations could be automated out; will not just affect blue-collar, but also professional/white-collar jobs. What’s left for humans to do?
- General, all-purpose jobs?
- Creative work? Personal, empathy-based work?
- Could it improve human society?
- AI could take over undesirable tasks, such as tedious or repetitive work
- Could automation of certain tasks free-up time and attention, allowing humans to focus on other areas? e.g. marking programs in education = more time devoted to students
- But also has material consequences; e.g. e-waste, etc.
- How do we transition?
- How do we support people who are losing these jobs? How do we retrain/educate?
- Universal Basic Income; leisure-based economy
- How do we convince society to accept changes like UBI?
- How do we advise students? What do employers want? More specific or more general? → Role of university?
- AI is already impacting work/labour
- Obsession with productivity/optimization
- Mental health consequences → dealing with uncertainty, volatility, lack of security
- Social expectations around training; hyper-certification
- Work as demanding more of our time/attention; no longer 9-5pm
Session 3: Diversity in AI
Need for More Diverse Voices in AI:
- How do we create more welcoming environments?
- Importance of allies/building support networks
- Is affirmative action the only/best solution?
- Educating about benefits of diversity; importance of designing and programming for diverse groups of people
- Diversity as encompassing a multiplicity of issues: accessibility, age, class, language, etc.
- Is there a way to test whether apps/tech is accessible?
- Digital Positivism?
- Who has access to these technologies?
- Need for a neutral, citizen-led watchdog/oversight group
Courses of Action for Academia:
- Interdisciplinarity:
- Importance of breaking down and deconstructing disciplinary boundaries; e.g. through restructuring/renaming
- How can universities better connect with the public/greater community?
- Commit to making resources more public/accessible
- Ensure that academics are publishing open-source; not publish in journals that have a paywall
- Public Organizations:
- Alberta Machine Intelligence (AMI): Alberta research group; also leads educational seminars on AI skills and literacy (https://www.amii.ca/amii-educates/)
- StartUp Edmonton
- Curriculum/Training:
- Could university provide more AI-centric interdisciplinary courses? E.g. interdisciplinary AI ethics course?
- UAlberta is also developing intro to machine-learning courses that explain how these algorithms work; what’s under the hood *Are these courses open to students from all **departments/faculties?
- Technology and the Future of Medicine; open to all graduate students
- The need for better ethics courses in STEM (and humanities); student consultation in developing courses
- Seminars that centre, not just on learning, but conversation/discussion could also be helpful.
- Student Organization
- Could we organize an AI-centric meeting group or space for students to discussion these issues? UAlberta’s new digital centre might be one option.
- But need to keep in mind the time investment/emotional labour, esp. for students
Session 4: Transparency/Openness
- From developer standpoint → What do public want to know?
- What is AI? Where does it already exist? What are its capacities/limits?
- Could we flag tools/apps that are collecting your data?
- AI as fourth gender, dedicated pronoun: how to make AI tangible, as opposed to abstract
- How and when do we grant rights to AI? Does AI have personhood?
- Definitions of human as exclusionary
- Is there again a risk of offloading responsibility from the designer? The collaborative nature of tech companies, with large research and development teams, boards of directors, etc., makes it *difficult to assign blame or make companies accountable
- Should developers be able to explain why certain outcomes/decisions are being made?
- Transparency around the dataset
- Do developers have an obligation to use datasets that reflect the population? *BUT it’s also noted that vulnerable populations are understandably wary of handing over their information. How *do you protect underrepresented groups?
- There is a tension between the need for privacy and anonymization and need for better, more representative data
Possible Action Outcomes:
- How to encourage more communication/collaboration between students in other disciplines? Could we organize a joint event between comp-sci/engineering and humanities, such as a workshop or Q&A on campus?
- Propose interdisciplinary course/workshop?
- Create an action Statement for the University that could embody set of goals?