Super interesting point about graphing calculators improving math scores - even when they weren't used. I'll counter by saying this seems to be very different than the effect of more recent tech, namely smartphones, which appear to be associated with "a growing struggle to concentrate, and declining verbal and numerical reasoning" (Financial Times, March 2025).
The question: Will AI be more like graphing calculators or smartphones? Your argument is that it's like a calculator, and given how passive most smartphone use is, versus how active you need to be to use AI well, I'm inclined to agree.
Final point: I'm not convinced AI "will keep getting smarter." At the end of the day, these are just very useful probabilistic models. Sticking with your calculator analogy, the iterations of GPT-3.5 to GPT-4o seem along the lines of TI-82 to TI-83. Incremental, and ultimately limited.
Just as Texas Instruments never became our Supreme Overlord, neither will ChatGPT (or Claude, or DeepSeek, etc., etc.).
Thanks Ryan for the note! I agree -- smartphones are more of a consumption device than our current iteration of AI interfaces.
I'm holding out hope that you will see AI getting smarter. Yes, they are probabilistic models, but the improvements in the last 6 months are quite impressive. GPT-4o was released a year ago next month (which is also insane given that it feels more recent.)
Super interesting point about graphing calculators improving math scores - even when they weren't used. I'll counter by saying this seems to be very different than the effect of more recent tech, namely smartphones, which appear to be associated with "a growing struggle to concentrate, and declining verbal and numerical reasoning" (Financial Times, March 2025).
The question: Will AI be more like graphing calculators or smartphones? Your argument is that it's like a calculator, and given how passive most smartphone use is, versus how active you need to be to use AI well, I'm inclined to agree.
Final point: I'm not convinced AI "will keep getting smarter." At the end of the day, these are just very useful probabilistic models. Sticking with your calculator analogy, the iterations of GPT-3.5 to GPT-4o seem along the lines of TI-82 to TI-83. Incremental, and ultimately limited.
Just as Texas Instruments never became our Supreme Overlord, neither will ChatGPT (or Claude, or DeepSeek, etc., etc.).
Thanks Ryan for the note! I agree -- smartphones are more of a consumption device than our current iteration of AI interfaces.
I'm holding out hope that you will see AI getting smarter. Yes, they are probabilistic models, but the improvements in the last 6 months are quite impressive. GPT-4o was released a year ago next month (which is also insane given that it feels more recent.)