Koko and animal “language”

Can animals have language the way humans have language? From recent news headlines about a famous ape and her passing, one might conclude: yes. Titles such as “Koko: Gorilla who mastered sign language dies in California” and “Koko, the Gorilla Who Knew Sign Language, Dies at 46” imply that non-human primates, at least, can learn and use language as humans do. So once again I’d like to call the media out on their frustrating inaccuracy.[1]

From the BBC (first article linked above):

“Koko the gorilla, who is said to have been able to communicate by using more than 1,000 hand signs, has died in California at the age of 46. Instructors taught her a version of American Sign Language and say she used it to convey thoughts and feelings. […] The gorilla […] could understand 2,000 words of spoken English.”

From NBC (older article linked to second one above):

“In 2001, Williams visited the Gorilla Foundation in Northern California, where he met Koko, a gorilla who is fluent in American sign language”

By choosing words like master and know and fluent, journalists are misinforming readers (whether this is intentional or out of ignorance, I’m not sure). There are many differences between how Koko understood and used language and how humans understand and use it. Two obvious areas jump to mind: (1) vocabulary size; and (2) ease of learning. Regarding point one – adult native English speakers know between 270% – 500% more words than Koko[2]. A 2016 study found that 20-year-olds were able to identify an average of 42,000 lemmas (dictionary headwords, e.g. run for run, runs, ran, and running). The lowest 5% of the population in the study recognized 27,100 lemmas, and the highest 5% understood 51,700.[3] While I couldn’t find numbers for American Sign Language (ASL) speakers, I assume they would not be too far from the numbers for speakers of English. (Most ASL speakers are bilingual anyway, since they have learned English to read and interact with non-signers.)

As for point two, the ease of learning – Koko had to be explicitly instructed over many years. Human children acquire a relatively complete grammar within about 5-8 years of being born, and they do this even in the absence of active teaching.

More generally, major properties of human language which distinguish it from animal communication are:

  1. Arbitrariness – the relationship between sounds/signs and their meanings is random.
    • Ex.: There’s no reason for why the sound sequence cat has to mean those furry purring domesticated felines.
  2. Productivity/Creativity – people can understand and create an unlimited number of new utterances.
    • Ex. 1: Novel sentences. I bet you’ve never heard the following, although you can easily understand its meaning.
      • Her pet squid squirted aquamarine ink all over the plaid sofa.
    • Ex. 2: Endless modifiers (adjectives, adverbs, etc.).
      • I thought he was really, really, really, really, really, very, super, exceedingly funny.
    • Ex. 3: Subordinate clauses (the below is one type, called a relative clause).
      • This is the dog that worried the cat that killed the rat that ate the malt that lay in the house that Jack built.
  3. Discreteness – human languages are composed of distinct units that combine according to grammatical rules to create meaning.
    • Ex.: Cat is made up of the phonemes /k/ + /æ/ + /t/; cats is made up of the morphemes cat + -s (pluralizes nouns); and on and on into larger and larger units.
  4. Displacement – we have the capacity to talk or sign about things unrelated to “the here and now” (the present space and time).
    • Ex. 1: When he was little he loved turtles.
    • Ex. 2: If I could do anything, I’d become an astronaut and travel to Mars.

Of course, animal communication has been shown to exhibit one or two of the above properties in specific contexts; you may have heard about such research on bird song, bee dances, and the communicative systems of elephants, bats, dolphins, squids, and apes. But at this point we don’t have evidence to suggest that any non-human form of communication displays all of these aspects simultaneously, or to the extent that human language does. That’s not to say that informational exchanges between other species aren’t complex. And this post is not meant to detract from Koko’s life or death. She and her trainers taught us invaluable things about the depth of gorilla intelligence and emotionality. Certainly, as we learn more about how various species communicate, we may need to update our definition of “language”. We’re just not there yet.

 

[1] The widely-read Language Log also posted on this topic, taking a similar stance: The (Non-) Evolution of language
[2] I won’t get into how “word” is defined, since that by itself would be a very long post.
[3] https://www.frontiersin.org/articles/10.3389/fpsyg.2016.01116/full

On machine translation, the media, and meaning (a response)

Shallowness_GoogTrans

I’m a Douglas Hofstadter fan. I read his book I Am a Strange Loop years ago, and it remains one of my three favorite non-fiction books, period. I highly recommend it to anyone who is at all interested in the nature of consciousness. The cognitive scientist’s Pulitzer Prize-winning Gödel, Escher, Bach: An Eternal Golden Braid has also been on my to-read list for a long time. So I was excited to see this article by him in The Atlantic, on another area that interests me: machine translation and machine “intelligence”.

Early on in the piece, Hofstadter says he has a “longstanding belief that it’s important to combat exaggerated claims about artificial intelligence”. Having worked in the machine learning/AI field for a little under a year now (but what an intense year it has been!), and having read countless popular media articles touting the astonishing advances in natural language processing/understanding, ML, and AI, I heartily agree with his sentiment. Such reporting is as misleading as it is annoying.

I came across a statement of this type the other day, in Stanford AI researchers make ‘socially inclusive’ NLP:

“The average person working with NLP today may consider language identification a solved problem.”

I have trouble believing that any researcher working in NLP/NLU/ML/AI thinks anything is a solved problem. Despite much progress, the field is still in its infancy. Doesn’t anyone remember Einstein’s quote (adapted from a similar idea expressed by Socrates) – “The more I learn, the more I realize how much I don’t know”? Where I work, every possible solution to a given problem brings up more questions, and even the “simplest” “facts” cannot always be taken for granted. (Remember when you were taught parts of speech like verb, noun, and preposition in grade school? Working at the level of detail we do, even these fundamental rules are often inadequate, requiring further specification. Turns out it’s hard to throw messy, real language into clean, fixed bins.) So I think the media does the field, its researchers, and the reading public a great disservice by sensationalizing and oversimplifying the challenges.

Hofstadter’s argument about understanding is even more poignant:

“The practical utility of Google Translate and similar technologies is undeniable, and probably it’s a good thing overall, but there is still something deeply lacking in the approach, which is conveyed by a single word: understanding. Machine translation has never focused on understanding language. Instead, the field has always tried to ‘decode’— to get away without worrying about what understanding and meaning are.”

We call the study of meaning and understanding semantics and pragmatics. People’s real world knowledge plays a key role here as well. To my mind, meaning (only complete when tied to real world knowledge) is the last frontier for AI and language. Today’s mobile/home voice assistants have definitely not yet mastered meaning. Technologies have made serious headway in resolving structural patterns (syntax), proper nouns (Named Entity Recognition) and some other aspects of language. But meaning, that great magical beast, eludes its pursuers. It is really, really challenging to computationally model the depth and complexity of human understanding. Because, although language itself is quite complicated, it’s still an impoverished medium for conveying the millions of subtle things we want and are able to convey – it relies heavily on context, implicature, presuppositionentailment, prosody, speaker-listener relationship, etc. I agree again with the author when he says that human-like machines are “not around the corner.”

I do think that Hofstadter seems to be simultaneously recognizing how hard the task of modeling meaning is, while not giving enough credit for things accomplished. Google Translate is way better now than it was at its inception over ten years ago. I also think he glosses over the tool’s main usage – which is more functional than artistic or poetic. Am I wrong to assume people use Translate much more for a quick word or phrase, when traveling or speaking on the fly, than for translating longer passages of literature? If they’re doing the latter… per the author’s experimental results, they clearly shouldn’t be.

 

What do you think – about media reportage of technology, machine “intelligence”, Hofstadter’s article? Feel free to comment!