Lisp was the goto languages for symbolic AI, but now that symbolic AI has been eclipsed by machine learning, Lisp is no longer the "language of AI" that it once was. Why did Lisp fail to make inroads into machine learning like it did for symbolic AI?
Because LISP does is not about machine learning.
>>1
Lisps (and Wirths too) fell out of use in academia generally following changes in industry.0 There are still some interesting Lisp solutions however if you're interested in investigating.1
0 https://cemerick.com/blog/2009/03/24/why-mit-now-uses-python-instead-of-scheme-for-its-undergraduate-cs-program.html
1 https://probprog.github.io/anglican/
There is no specific reason, it's just how history turned out. There's nothing in Python that makes it any good for machine learning. All the heavy lifting is usually done by calling non-Python code anyway. Any other language would have done it, but Python was the most trendy at the time.
>>4
A model that assumes some event is random is more indicative of ignorance (or if you like uncertainty) than of the event itself.
>>5
Lisp was popular for symbolic AI because it has language features (notably symbols) that make it particularly suitable for it. This is not the case for Python for machine learning, which does not have anything that would make machine learning particularly suitable for it. In fact, if machine learning workloads were written in pure Python, they would have been a disaster as Python is extremely slow and these workloads are very computation heavy. Python became so strongly associated with machine learning because the popular early frameworks and libraries were written for it, and they were written for it because Python was and still is a popular language. If some Lisp was popular at the time there's no reason it couldn't have become the language of machine learning.
>>6 I agree.
Is there still any research into symbolic artificial intelligence? It sounds so much more exciting than the current hyped statistical predictors.
Because silly commonlispers haven't jumped onto the clojure ai hypetrain
/thread
>>8
Sure, it just doesn't get the same publicity, and is more boring because you need to learn (mathematical) logic first.
Are there journals or conferences for symbolic AI research? I want to see what they are up to, but have no idea where to start.
Lisp (and Prolog) were/are popular for Symbolic AI because they feature symbols as a first-class data type. With Machine Learning, AI became more about numerical analysis, which can be efficiently done in more mundane languages like C. Hence, Lisp (and Prolog) lost their main advantage for AI and subsequently fell into the sidelines.
if self.rear is None:
self.front = node
self.rear = node
else:
self.rear.next = node
self.rear = node
My intelligence is symbolic, too: it's there, but it is not suitable for practical use.
Heard that.