Merge pull request #71 from ContextualAI/nina/aie-workshop

update eval_input file

Author: ninalopatina

08-ai-workshop/data/eval_input.csv

@@ -0,0 +1,59 @@
+prompt,response
+What was NVIDIA's Data Center revenue in Q4 FY25?,"NVIDIA's Data Center revenue for Q4 FY25 was $35,580 million.[1]()[2]()
+
+This represents the largest segment of NVIDIA's total revenue for the quarter, which was $39,331 million.[1]()[2]()"
+What is the correlation coefficient between Neptune's distance from the Sun and US burglary rates?,"The correlation coefficient (r) between Neptune's distance from the Sun and US burglary rates from 1985-2022 is 0.969, with r²=0.940 and p<0.01.[1]()
+
+However, it's crucial to understand the context of this correlation:
+
+This analysis involves several important limitations:
+- It uses a method called 'data dredging' where 25,153 variables were compared against each other to find matching patterns[1]()
+- The analysis lacks a direct causal connection between the variables[1]()
+- The observations are not independent, as sequential years are connected[1]()
+- The visual representation is intentionally misleading because the Y-axis does not start at zero[1]()
+
+Given these limitations, while the statistical correlation exists, it should not be interpreted as a meaningful causal relationship between Neptune's position and burglary rates."
+How did NVIDIA's total revenue change from Q1 FY22 to Q4 FY25?,"I can provide NVIDIA's total revenue for Q1 FY22 and Q4 FY25, though I don't have data for Q4 FY25's comparison to Q1 FY22 directly. Here are the relevant figures:
+
+NVIDIA's total revenue in Q1 FY22 was $5,661 million[3]()
+
+NVIDIA's total revenue in Q4 FY25 was $39,331 million[1]()
+
+This represents a significant increase from Q1 FY22 to Q4 FY25, though for a precise quarterly comparison, you would need to compare Q1 FY22 to Q1 FY25. Based on the available data:
+
+The Q1 FY25 total revenue was $26,044 million[1]()
+
+To calculate the exact change from Q1 FY22 to Q4 FY25, you would need to consider the entire period, but the data shows a clear upward trend in revenue over this timeframe."
+"What are the four main reasons why spurious correlations work, according to the Tyler Vigen documents?","Based on the provided documentation, here are the four main reasons why spurious correlations can occur:
+
+1. Data Dredging: This involves comparing large numbers of variables against each other to find random matches. With 25,153 variables, this results in 632,673,409 correlation calculations. Any sufficiently large dataset will yield strong correlations completely at random.[1]()
+
+2. Lack of Causal Connection: There is typically no direct connection between the variables being compared. This is particularly problematic when using years as the base variable, since many unrelated things happen within a year.[1]()
+
+3. Observations Not Independent: For many variables, sequential years are not independent of each other. When people are continuously doing something every day, there's no reason to expect sudden changes just because the calendar changes to a new year. Standard p-value calculations don't account for this, making the correlations appear less probable than they actually are.[1]()
+
+4. Manipulated Y-axis: When the Y-axis doesn't start at zero, it can create a misleading visual impression. This is intentionally used to make the correlation appear more significant than it actually is.[1]()
+
+These factors can combine to create seemingly impressive but actually meaningless correlations, as demonstrated in the examples provided in the documentation."
+Why should we be skeptical of the correlation between Unilever's revenue and Google searches for 'lost my wallet'?,"There are several compelling reasons to be skeptical of this correlation between Unilever's revenue and Google searches for 'lost my wallet':
+
+1. Data Dredging: This correlation emerged from analyzing 25,153 variables, resulting in 632,673,409 correlation calculations. This approach, known as data dredging, can produce strong correlations completely at random in large datasets.[1]()
+
+2. Lack of Causal Connection: There is likely no direct connection between these variables. The analysis used years as the base variable, which means many unrelated things could be happening simultaneously.[1]()
+
+3. Non-Independent Observations: The sequential yearly observations are not independent of each other. Human behavior doesn't suddenly change on January 1, which affects the statistical validity of the findings.[1]()
+
+4. Visual Manipulation: The graph's Y-axis doesn't start at zero, and a line graph was used, which intentionally makes the visual connection appear more significant than it actually is.[1]()
+
+While the statistical measures might look impressive:
+
+- The correlation coefficient is 0.921[6]()
+- The p-value is 4.005E-7[6]()
+- This level of correlation would occur by chance in only 4.005E-5% of random cases[6]()
+
+These statistical measures are misleading due to the methodological issues mentioned above. The correlation, while mathematically real, is likely spurious rather than meaningful."
+When did NVIDIA’s data center revenue overtake gaming revenue?,"Based on the provided data, I can identify when NVIDIA's data center revenue first exceeded gaming revenue, though the exact quarter of initial crossover isn't directly visible in the data provided:
+
+In Q1 FY24, Data Center revenue was $4,284 million, while Gaming revenue was $2,240 million. This represents the first quarter in the provided data where Data Center revenue clearly exceeded Gaming revenue.[1]()
+
+However, to determine the exact quarter when this transition first occurred, we would need data from earlier periods not included in the provided information."