@Jonas
AI progress and alignment studies
jonas$0 in pending offers
I am Jonas, the founder of the African AI Safety Lab, Equiano Institute, and Progress Studies Journal. I am passionate about promoting AI alignment and policy for African. I pursued my studies in Computer Science, earning a candidate for Masters at the University of Sussex and completing a BSc in Mathematics at Istanbul University. As an independent researcher, I established the Progress Studies Journal, which serves as an overlay journal dedicated to distilling research aimed at technological progress, and responsible innovation. Previously, I served as Founder in Residence at Enterpreneur UK in London, where I worked with a select group of entrepreneurs to build projects and foster partnerships. I am also actively involved with Cohere For AI as a volunteer, leading AI Safety Talks. In this role, under Sara Hooker.
Projects
Comments
Jonas Kgomo
6 months ago
What progress have you made since your last update?
The Equiano Institute is a responsible AI research lab for Africa and the Global South. Here are our updates for the past 6 months:
Capacity Building for UN Diplomats
We have assisted the UN in preparation towards the Global Digital Compact, through consulting and training UN Diplomats. We also consult for the UNDP, UNICEF etc on issues related to AI in Africa. We are contributing members of the Harvard Ash Center research network on Getting Plurality.
Capacity
We have run a Fellowship on AI governance and technical alignment.
Research Agendas:
Policy Impacts of AI: We publish research on AI policy development and technical safeguarding of digital public infrastructure.
Economic Impacts of AI: How does AI affect individuals and communities in the Low and Middle Income Countries
Open-Source Projects
We are currently supporting OpenAI in developing multilingual language models for improving AI performance in African Languages.
Coalitions
We currently spearhead coalition initiatives through two working groups.
The AI Working Group directly addresses challenges faced in the region by promoting responsible AI practices and tackling societal implications, focusing on issues like data access, governance, and participation in AI development.
The Africa Oversight (TAO) is a technically-focused initiative focused specifically on ensuring the safe and ethical development of AI technologies in Africa. TAO utilizes multifaceted evaluation methods and aims to develop a comprehensive framework for best practices in ethical AI development and deployment within the African context.
The goal is to promote transparency and accountability in the use of artificial intelligence (AI) in the public sector.
These combined efforts address the specific needs of Africa’s responsible AI trajectory.
Thank you for your time and consideration.We have published peer-reviewed papers in conferences such as Data for Policy, NeurIPS and others.
What are your next steps?
Global AI Benchmarks
We are building a more Plural MMLU (this is a normative approach to evaluations that captures contextual differences between nuances in different languages and cultures).
Government DPI AI Safety
We believe that our project on multilingual African language can be a core part of government services, as various governments integrate AI into their government infrastructure as we have seen recently:
Nigeria launched first multilingual LLM
Microsoft in Kenya launched their digital ecosystem initiative for Kenya which focuses on Swahili/English AI models and AI societal services
Research Themes
We have research themes on Health, Development and AI this programme looks at the intersection of explainable AI in the development sector. Research we are excited about:
Development Impacts: We have research themes on Health, Development and AI. This programme looks at the intersection of explainable AI in the development sector.
Social Impacts: this considers how public good AI interse normative and social aspects of AI Readiness/Strategy Frameworks
Is there anything others could help you with?
Funding for researchers contributing to our projects. Currently all our research is on volunteer basis.
Research collaborations are welcome
Jonas Kgomo
about 1 year ago
Some updates after 30 days of launching the project without any funds:
- we have more than 40 researchers, including a few high school students interested in ai governance and alignment
- 5 research projects with an average count of 3 -10 researchers per project (researchers are mostly from META, MIT, Harvard, Open AI and Cornell).
- we have been timely since African nation states are getting keen about AI reception in Africa
- collaboration with GovAI, BlueDot on the fellowship
- organising a zero-cost online fellowship
- we are submitting our first paper on Factored Cognition of LLMs using causal influence diagrams,
- we have world-class PhD mentors leading projects at Equiano Institute
I would like to tag people who have regranted us before Manifund: @LeopoldAschenbrenner @LinchuanZhang
And people we have engaged with in the past few weeks @RenanAraujo
Thank you, would appreciate any thoughts.
@Austin
Jonas Kgomo
about 1 year ago
pinging to share with you on this final day pre-archival
@joel_bkr @GavrielK @IsaakFreeman @AntonMakiievskyi
Jonas Kgomo
about 1 year ago
I can say this organisation is likely to have impact as we are also considering working with them on a fellowship
Jonas Kgomo
over 1 year ago
Agreed, we have seen different failure modes and talent agglomeration that are unique to Africa and that might be absent in other AGI hubs. The training and capacity building is a special and main part of the project, we want to launch a broader talent search program. So far we have been doing targeted outreach and community building, which is less intensive than launching a lab. We are using the approach of grassroots research lab, which has proven to be effective in Africa for projects like Masakhane NLP, Cohere For AI, Deep Learning Indaba [https://www.youtube.com/watch?v=qSCQ2Tv3YmA]
We have seen limits to global inclusion in AI development [https://arxiv.org/abs/2102.01265] , hence opinions represented by the models are less similar to participants in Africa [https://llmglobalvalues.anthropic.com] . We aim to have an interdisciplinary community capable of advising on AI governance and work on technical AI alignment research (especially for reducing disparities, bias and suffering "s-risk" caused by AI)
Transactions
For | Date | Type | Amount |
---|---|---|---|
Manifund Bank | about 1 year ago | withdraw | 1500 |
Manifund Bank | about 1 year ago | withdraw | 100 |
AI Alignment Research Lab for Africa | about 1 year ago | project donation | +1500 |
AI Alignment Research Lab for Africa | about 1 year ago | project donation | +100 |
Manifund Bank | about 1 year ago | withdraw | 850 |
AI Alignment Research Lab for Africa | about 1 year ago | project donation | +200 |
AI Alignment Research Lab for Africa | about 1 year ago | project donation | +150 |
AI Alignment Research Lab for Africa | about 1 year ago | project donation | +100 |
AI Alignment Research Lab for Africa | about 1 year ago | project donation | +150 |
AI Alignment Research Lab for Africa | about 1 year ago | project donation | +250 |