Marco Argenti, chief information officer at Goldman Sachs, is advocating for his team to delve into engineering research to better deal with the complexities of AI technology.
Mr. Argenti emphasized the importance of critical thinking skills and encouraged his staff (and his daughter) to explore fields such as philosophy alongside traditional engineering.
In recent articles, harvard business reviewMr. Argenti emphasized the value of philosophical research, citing its role in honing problem-solving skills.
While knowledge of Aristotle, Plato, and Socrates alone won't secure you a position as an AI engineer, Argenti believes that when combined with technical expertise, it can improve the quality of your code.
He explained: “The ability to understand a problem by developing a clear mental model around the problem you want to solve. why Before starting work, how This is an increasingly important skill, especially in the age of AI. ”
Indeed, while aspects of AI have been incorporated into technology services for years, the launch of large-scale language models (LLMs) such as ChatGPT and Microsoft's Bing bot has sparked a wave of new product launches as rivals try to catch up. is surging.
But most are doomed to failure, at least according to former Harvard University professor Clayton Christensen, who said that 95% of new services launched in any field will not succeed.
And while Argenti said AI can “write higher quality code than humans,” there is a catch: “Sometimes it works well, but it doesn't do what you want it to do.”
This is where the technique of prompt engineering comes in, writes Argenti. “The quality of the LLM output is highly influenced by the quality of the prompts. Ambiguous or ill-formed questions will cause the AI to try to guess what question the user is actually asking, and as a result, You're more likely to get an answer that's inaccurate or completely fabricated.”
Perhaps it's no surprise that, as of this writing, salaries for these roles range from $30 per hour to $405,000 per year.
Landing one of these roles and asking questions to train an AI to work as intended requires “reasoning, logic, and first-principles thinking” and requires “all the basic The skills were developed through philosophical training,” continues Argenti.
Revitalizing philosophy
For years, philosophy students have had to accept that their degree usually won't land them a well-paying job.
Education platform HeyTutor cites an analysis by the Federal Reserve Bank of New York's US Census Bureau that philosophy departments have one of the highest post-graduation unemployment rates in the US at 6.2%.
The underemployment rate for this sample is just over 50%, and the median early career wage is $36,000.
But with support from Argenti and others, things could change.
He writes: “Having a clear mental model of a problem, being able to break it down into manageable steps, and being ready (and able) to argue for perfect first-principles thinking, sometimes stubborn AI. can make a good engineer, and the same considerations probably apply to many professions.”
And AI has indeed proven to be stubborn at the best of times.
Take Microsoft's Bing bot as an example. The bot reportedly gave one user the wrong date and then reprimanded the user for pointing out the bot's mistake. It then told the user to apologize and instructed them to “end this conversation and start a new one with a better attitude.”
Argenti added that it is essential not to lose the skill to “pop the hood” of large-scale language models, but added: It has a disproportionate impact on the world.
“AI makes us less like computers and more like humans.”