First, for next year's Leadership in Action, I hope I can work for an AI lab on the character and impact of AI somewhere outside the UK, ideally US.

For our database, it is true that AI is not that reliable when using its web search function to fill in the blanks of Excel sheets. Therefore, we finally decided to produce two versions:

1. One version where we only use APIs from trustable secondary sources.
2. A second version where we use the ChatGPT API to fill in those blanks.

We are still working to finish both and compare the differences between them.

Meanwhile, I am also responsible for drawing findings from the data we already have, though we have realized it is quite hard to find significant findings. For example, we expected that factors like the distribution of sectors in different locations and their export shares would be different, but it actually doesn't surpass a standard deviation. We are also trying to find correlations between different statistics, but still have no significant findings.

Though they are tangential findings, our supervisor, Mon Chen, suggested some useful perspectives:
• By looking at the year the CIO form came into place, we can see a spike in newly registered CIO companies.
• We noticed a decline in newly established companies in 2009.

If I could do this differently and start from the beginning, we might choose to iterate more frequently. The traditional form of working in the pre-AI era was to first set a timeline for the project and then divide the work among each member of the team. But now, in this AI era, in my opinion, the determining factor of the quality of a project is moving from the consistency of your work to its overarching structure and design. Therefore, sometimes it is easier to completely abandon our initial ideas and intermediate results and start from the beginning. In other words, we should iterate instead of editing again and again.