Suggestions
Arthur Berd
Quantitative Investment Management, Financial Planning and Wealth Management
Arthur Berd is a highly accomplished professional with a strong background in quantitative finance and technology. He has an impressive academic and professional history:
Educational Background:
- Ph.D. in Physics from Stanford University
- Ph.D. program in Theoretical Astrophysics at the Landau Institute for Theoretical Physics
- M.S. in Theoretical Astrophysics from the Lebedev Institute of Physics
- M.S. in Theoretical Physics from the Moscow Institute of Physics and Technology
Professional Experience:
- Current Advisor at TicketJan1
- Previously Head of Advice at MoneyLion (2020-2022)15
- Founder of Wealth Technologies Inc., which was acquired by MoneyLion in 20201
- Founder and CEO of General Quantitative, LLC (2012-2020)15
- Head of Macro Vol Strategies at Capital Fund Management (2008-2011)1
- Head of Quantitative Market Strategies at BlueMountain Capital Management (2005-2007)1
Berd is based in Miami, Florida, and specializes in Quantitative Finance, Systematic Investment Management, and Algorithmic Trading.234
Highlights
I have two related hypotheses to explain this degradation on ChatGPT abilities:
-
OpenAI has introduced the human feedback and learning from interactions into the system and it has been “learning” from all the inane questions that everyone has been asking, trying to trick it into giving silly answers.
-
OpenAI has made it “personalized” and, like a tired scientist in a bad Twitter thread, it just brushes off all questions assuming that they are silly unless the question is asked in such a manner as to make it clear that the person is indeed serious and is worthy of a serious answer. In other words only sophisticated prompts receive proper answers, while the rest receive lower quality random feedback that is designed to measure the questioner’s response and performance rather than make an effort to make sense of the question.
The first hypothesis is partly in jest, but the second is a plausible approach to keep the technology growing without freaking out all the AI doomsayers. Or perhaps it’s just an attempt to make the free version of the technology less capable in order to be able to sell more capable versions by enterprise subscription?
Completely agree with Terence Tao’s comment - whether it is a homework, exam or interview, the best mode of testing should not be a test of “knowing” but a test of “understanding”.
