Building safety controls into AI systems
Washington, DC
Joined February 2014
- Tweets6.8K
- Following710
- Followers800
- Likes2K
Pinned Tweet
This extinction or catastrophe mechanism should be front and center of any global AI governance effort as we look forward to the U.S.-China summit this week.
BREAKING: The US "almost started a war" against China, nearly launching an operation against a Chinese ship in the Middle East this spring, in the midst of the Iran war, after an AI-generated intelligence report falsely identified the vessel as carrying nuclear weapons components, with armed US personnel preparing to board and military aircraft already in the air before officials found the report had been produced with a chatbot, per CNN.
A special operations command analyst had used a chatbot to fuse open-source material with classified signals intelligence, then used AI again to package the result into a standard intelligence report and circulated it across the US military.
Michael Akinwumi retweeted
In 2025, local fair housing orgs and the DOJ processed 20,805 housing discrimination complaints, 55% involving a disability.
Read NFHA’s 2026 Fair Housing Trends Report to learn how we are advancing Fair Housing alongside our member organizations: vist.ly/5qisw
ALT Infographic with text about 20,805 housing discrimination complaints in 2025.
ALT In 2025, 55% of reported complaints involved disability.
ALT Text overlay on purple background reads: 'Processed complaints declined 14% compared to 2024. Lower numbers don’t mean fewer acts of illegal discrimination.'
ALT Purple infographic with text about missing data and funding disruptions in housing.
A best is a resting place, not a destination. From every peak, the view reveals a higher one. When did you last believe you had arrived? What did that belief cost you?
Michael Akinwumi retweeted
Funding disruptions are leaving communities without local fair housing to help.
NFHA's 2026 Fair Housing Trends Report found three out of four reporting member organizations faced funding disruptions.
Full report: nationalfairhousing.org/reso…
Who does the learning when AI is in the classroom?
I’m moderating that conversation at the 11th OneGW Summit with a student and @gwbusiness' Chief AI Officer and Teaching Assistant Professor of Decision Sciences, bring the student-faculty views. I will also chip in social metacognition, thinking together about how we think, with and without AI.
Who evaluates the machine? In one influential AI essay, the answer is: the firm, against its own business outcomes, in private.
Until the many can read the scorecard, the house is still grading itself.
New essay: datawumi.substack.com/p/the-…
ALT Until the many can read the scorecard, the house is still grading itself.
Microsoft is reportedly offering buyouts to 60,000 employees while spending up to $175bn on AI infrastructure this year.
The report doesn't say one caused the other, and voluntary buyouts are not layoffs. But the timing raises the question my latest essay asks of the whole ecosystem: as firms build "token capital," what happens to human capital, and who has a say in that trade-off?
It's a question for every large employer, not one company.
Michael Akinwumi retweeted
A 2026 Fields Medalist has made a surprising admission about AI.
John Pardon says he has used AI to find “a few nontrivial proofs” for mathematical results he believed were true but hadn't fully established.
AI has also caught subtle errors in his arguments.
But there is a catch: it still struggles with some of his most serious research questions.
His latest interview offers a glimpse into how one of the world's leading mathematicians actually works with AI.
Michael Akinwumi retweeted
Fermi to Dirac, 100 years ago today: You didn't cite my paper. ;P
(h/t Physics Archives: facebook.com/photo?fbid=1073…)
My undergraduates are learning decision trees, random forests, and gradient boosting through Andrew Martin's model of SCOTUS decisions.
This headline suggests political analytics is a viable career.
Michael Akinwumi retweeted
Black Population by State in the US:
1. 🇺🇸 Mississippi - 38.49%
2. 🇺🇸 Louisiana - 33.22%
3. 🇺🇸 Georgia - 32.29%
4. 🇺🇸 Maryland - 31.72%
5. 🇺🇸 Alabama - 26.76%
6. 🇺🇸 South Carolina - 25.5%
7. 🇺🇸 Delaware - 23.66%
8. 🇺🇸 North Carolina - 21.61%
9. 🇺🇸 Virginia - 20.76%
10. 🇺🇸 New York - 17.6%
11. 🇺🇸 Tennessee - 16.84%
12. 🇺🇸 Florida - 16.36%
13. 🇺🇸 Arkansas - 16.02%
14. 🇺🇸 Illinois - 15.35%
15. 🇺🇸 Michigan - 15.18%
16. 🇺🇸 New Jersey - 14.8%
17. 🇺🇸 Ohio - 14.5%
18. 🇺🇸 Connecticut - 13.22%
19. 🇺🇸 Texas - 13.02%
20. 🇺🇸 Pennsylvania - 12.84%
21. 🇺🇸 Missouri - 12.64%
22. 🇺🇸 Nevada - 11.32%
23. 🇺🇸 Indiana - 11%
24. 🇺🇸 Massachusetts - 9.75%
25. 🇺🇸 Rhode Island - 9.5%
26. 🇺🇸 Kentucky - 9.49%
27. 🇺🇸 Oklahoma - 9.33%
28. 🇺🇸 Minnesota - 8.5%
29. 🇺🇸 Wisconsin - 7.51%
30. 🇺🇸 Kansas - 7.48%
31. 🇺🇸 California - 7.16%
32. 🇺🇸 Nebraska - 6.3%
33. 🇺🇸 Arizona - 6.08%
34. 🇺🇸 Washington - 5.79%
35. 🇺🇸 Colorado - 5.48%
36. 🇺🇸 Iowa - 5.27%
37. 🇺🇸 Alaska - 5.17%
38. 🇺🇸 West Virginia - 4.86%
39. 🇺🇸 North Dakota - 4.11%
40. 🇺🇸 Hawaii - 3.68%
41. 🇺🇸 New Mexico - 3.36%
42. 🇺🇸 Oregon - 3.21%
43. 🇺🇸 South Dakota - 3.13%
44. 🇺🇸 New Hampshire - 2.55%
45. 🇺🇸 Maine - 2.45%
46. 🇺🇸 Vermont - 2.23%
47. 🇺🇸 Utah - 1.95%
48. 🇺🇸 Wyoming - 1.65%
49. 🇺🇸 Idaho - 1.31%
50. 🇺🇸 Montana - 1.12%
Source: World Population Review
“Mathematically, an infinite supply of all goods and services would indeed make everything free. … Practically, however, such a cornucopia is almost inconceivable.” Yet Elon Musk, the world’s wealthiest man, believes in AI-induced abundance. @grok explain the contradictions in Elon’s paradox.
What “an uncomfortable question about ai-powered cyber-security: how can defenders get fixes to users, without tipping off enemies in the process?”!
Another Agentic AI seems to be a reasonable answer.
@grok isn’t that right?
Michael Akinwumi retweeted
Two critical vulnerabilities have been identified in PHP, a widely used programming language that powers websites and web applications. These vulnerabilities (CVE-2026-109618 and CVE-2026-109619) could allow malicious cyber actors to remotely execute code in PHP core, potentially compromising sensitive data and networks.
The FBI urges organizations to take the following steps now:
➡️ If an immediate update is not feasible it is recommended that organizations take the system offline until it can be fully mitigated.
➡️ Immediately update affected PHP installations following the guidance on php.net. Inventory where PHP runs in your environment, prioritizing internet-facing systems and vendor products that bundle PHP.
➡️ Monitor web servers for suspicious activity, including PHP processes unexpectedly launching command shells or downloading files. Ensure appropriate process-monitoring capabilities are in place to help identify potential exploitation.
➡️ If you find signs of compromise, isolate affected systems, preserve your web server and PHP logs, and report it to your local FBI field office or ic3.gov. Those reports help us identify the actors involved and warn other organizations before they become victims.
For the latest information, follow PHP’s announcement on php.net, CERT-EU Security Advisory 2026-018, and CISA’s Known Exploited Vulnerabilities catalog.
ALT Critical vulnerability alert for PHP Core with CVE-2026-109618 and CVE-2026-109619, CVSS score 9.2, urging immediate patch and threat hunting.
Michael Akinwumi retweeted
MIT announces the MIT for America initiative. The effort aims to help U.S. learners from kindergarten to community college, with an emphasis on math, making, and the constructive use of AI. news.mit.edu/2026/mit-americ…
Michael Akinwumi retweeted
OpenAI’s new proof of the quasi-Riemann hypothesis is, as Jared Duker Lichtman of Stanford University put it, “the most significant AI-for-math achievement so far, by quite a margin.”
quantamagazine.org/ai-proved…
Michael Akinwumi retweeted
"Mathematical Theory of the Transformer" is an extensive, 825-page mathematical monograph exploring the theoretical foundations of Transformer architectures.
It covers a remarkable range of topics, including embeddings and positional encoding, self-attention and multi-head attention, matrix decompositions, linear algebra, spectral analysis, information geometry, training dynamics, redundancy, optimization, and incremental Transformer architectures.
This is an advanced, mathematically rigorous read for anyone interested in understanding the mathematical structures underlying Transformer-based AI systems, including modern generative AI.
doi.org/10.5281/zenodo.20199…
Albertans spoke last night. The world now watches to see what the province and its neighbor to the south make of the referendum’s outcome.
Michael Akinwumi retweeted
Mathematics algorithm SVM (Support Vector Machine) is a classic nonlinear classifier in Machine Learning.
Download 640-page PDF "Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond" here: mcube.lab.nycu.edu.tw/~cfung…
