ELIZA Reinterpreted: The world's first chatbot was not intended as a chatbot | Hacker News Reader

ELIZA Reinterpreted: The world's first chatbot was not intended as a chatbot

arxiv.org

75 pointsmtraven24 comments

Except it WAS intended as a chatbot... as a tool "for research into human-machine conversation and the important cognitive processes of interpretation and misinterpretation".

It's like saying the guy who invented the screwdriver didn't invent it because he was REALLY just trying to build a cabinet. The reason a tool was invented does not mean the tool isn't what the tool is.

POSIWID.

It at least establishes that it wasn't intended as a public chatbot. To extend your analogy: it's as if the inventor of the screwdriver threw away his tool afterwards and someone found it in his trash and mass manufactured it. The cabinet maker wasn't trying to make a [new standard tool for how to drive screws] (screwdriver) but was making a tool [to drive screws for just one project] (screwdriver).

"the world first chatbot was not intended to be public."

More accurate, less interesting.

The “public” distinction feels anachronistic when applied to early 1960s software.

This was a time before networks and software distributions. The idea that software could be standardized and shared instead of written for a particular computer was not widespread.

So it’s a bit disingenuous to say that Eliza wasn’t intended as a public chatbot, when practically no software was intended to be public.

Yeah. That's actually one of the things that makes this story so interesting: It took place exactly at the dawn of the inter(arap)net, and physically close to BBN (which was the arpanet implementor), so there was lots of sharing between MIT and BBN, in particular, McCarthy put a lisp on BBN's PDP-1, and Bernie Cosell's Lisp ELIZA, which he did from the algorithm in Wiezenbaum's CACM paper, was the one that became public, so everyone thought that ELIZA had originally been implemented in Lisp, which was wrong!

But .... it sounds like it was was? The original MIT SLIP version (yes SLIP not LISP) had a framework for embedding dynamic modules - so could do an early version of "learning." It feels like it was indeed more a framework for testing out different kinds of chatbot logic - maybe a meta-chatbot?

The famous DOCTOR script was just one example to demonstrate the tools's capabilities, and not the main point of ELIZA.

[Edit: the paper suggests the DOCTOR script is "ELIZA" not ELIZA, i.e. the limited version that became popular, not the full version that was built)

It was intended as a chatbot in the sense that it provides an appearance of conversation when it’s basically a simple parlor trick. The technology of ELIZA wasn’t the interesting part, it was how people convinced themselves it actually had intelligence.

Jeff is a HN user and his connection to ELIZA is pretty interesting. Many of the ELIZA chat bot clones were based on a version he wrote as a kid, that was published.

He and a group of researchers put effort into digging up the history of original implementation.

Interesting person!

You can plug your interview.

I don't want to be that person, who's always plugging stuff when it doesn't fit. But yeah, here it is:

https://corecursive.com/eliza-with-jeff-shrager/

I think Jeff is also working on a book.

There's no way it doesn't fit.

Classic chatbots:

https://www.cs.cmu.edu/afs/cs/project/ai-repository/ai/areas...

Splotch it's offensive but fun.

Azile might run under Executor (the forked one at GitHub).

Emacs still has M-x doctor.

Probably based on Jeff Shrager's ELIZA. IIRC from https://corecursive.com/eliza-with-jeff-shrager/ most ELIZA's out there are actually based on his reimplementation, not on Weizenbaum's original code

Quick plug; mine is based on Weizenbaum's original (or at least on his notes/specifications, as the source code wasn't available when I built mine) https://christopherdrum.itch.io/eliza8

Is your source code open somewhere it can be read?

There is a link right on that very page to download, "eliza8_source.zip" Additionally, the "cartridge" can be opened by Pico-8 and viewed/edited/copied freely (free educational version of Pico-8 can do this as well https://www.pico-8-edu.com/)

And M-x yow. And a conversation between the two with M-x psychoanalyze-pinhead

https://sites.google.com/view/elizagen-org/

This is just word games isn't it?

Is arxiv just a blogging platform now? This isn't a research paper.

I think you misunderstand arxiv, blogging, and research papers.

(I've used/done extensively the three mentioned things before, including blogging in a research context)

arxiv is a preprint server. Blog posts can, sometimes, play the same role as research preprints. Research papers are fundamentally about having a _structured conversation_ about a topic.

This paper is arxiv at its best.

You should see some of Schmidhuber's stuff, he uses scientific journals as his blogging host.

Reply on news.ycombinator.com