A neuron as traditionally defined in biological terms is a basic unit of the nervous system which is responsible for broadcasting signals throughout the brain and body, providing us with agency over ourselves and the world around us (after all this is a marketing newsletter and so to incorporate the words ‘broadcasting’ and ‘agency’ in a definition of ‘neurons’, I thought would be cool). But in this future, we’ve suddenly found ourselves in, its definition is no longer necessarily relegated to a biological construct. A neuron is a connection point, or node, capable of processing incoming and outgoing information to other nodes. When bundled into a densely layered architecture neural networks can be anything which adds together values of inputs and spits them out another end as an output greater than the sum of its input parts. The human brain has 100 billion neurons, making for 100 trillion synaptic connections (the junction or site of transmission of electric nerve impulses between any 2 neurons). It is in us, what constitutes intention, thought and imagination.
GPT-4 will have (drum roll) 100 trillion neurons, as many as the human brain has synapsis; 500 times larger than its current phenom little brother ChatGPT-3. All indications are that we won’t even be able to test its full potential. There will not be a question you can ask it that it can’t answer better than any human on the planet. Will we create something so intelligent that it will have no use for us?
If you saw the recent headline in the NY Times “A.I. Poses Risk of Extinction, Industry Leaders Warn” you might think so! It says that AI could pose an existential threat to humanity, a risk on par with pandemics and nuclear war. Woah!! Whatcha talkin bout Wilis? I don’t know about you, but only two things scare me, and one is nuclear war. The other is not a smart aleck computer that I can unplug from the wall when I’ve had enough of losing arguments. Before we all go running through the streets like chicken little, it’s important to note that raw computing power does not make something conscious. What makes something conscious is it’s ability to ponder whether something it created is but doesn’t know because it’s not sure what that means to begin with. In other words, a self-aware being perplexed by its own self-awareness is a good proxy for consciousness. We can train a computer to mimic being self-aware but that doesn’t mean it is. By training GPT and other LLM’s on all the available knowledge that humanity has ever extracted and documented, it’ll do some serious dot connecting and be more like us staring back at our reflection in some fun house mirror than some hideously intelligent lifeform that poses an existential threat.
Alas, much further down the article the source of worry is revealed. LLM’s like ChatGPT are raising fears that “AI could be used at scale to spread misinformation or propaganda”. So, the headline grabbing hysteria notwithstanding, what’s worth considering is not AI becoming a sentient being but potentially being misused by one. Like a sword, a beautiful tool…a dangerous weapon.
On the Katz Television insights page you will find a study by the Pew Research foundation confirming that Mass Media, Local TV News in particular, is basically the glue that binds our society. It is the most trusted source of information and, as cited by the Knight Foundation, “local news media are a cornerstone of democracy…and foster a sense of community among their readers and viewers.” In an ambiguous world already replete with misinformation, the concerns over LLM’s making it worse should only stand to turn that “glue” that binds from Elmer’s to Krazy.
What LLM’s like Chat-GPT are already illuminating for us is that there is a template for all of language and that by reverse engineering the structure of the syntax we’ve come up with to convey thought we’ll have insight into a deeper understanding of how reality is organized. John Ambrose Fleming was a professor of electrical engineering and expert in the amplification and distribution of electrical power. In 1899 he was hired by Marconi’s Wireless Telegraphy Co. as Scientific Advisor. After a successful test transmission of radio waves including messages in English and French from a tower to a receiving station 30 miles away, Fleming sent a letter to The Times of London describing the experiment. “When it is realized that these visible dots and dashes are the result of trains of intermingled electric waves rushing with the speed of light across the intervening 30 miles, caught on one and the same short aerial wire and disentangled and sorted out automatically by the two machines into intelligible messages in different languages, the wonder of it all cannot but strike the mind”. You see, in the 19th century (when the test was conducted) drawing forth the mystical phenomena of electricity and magnetism was like pulling back a veil on reality which revealed an incomprehensible phenomenon that people could hardly believe. The implications on how it would eventually shape our society were unimaginable. Get ready. LLM’s and AGI are going to give form to a reality that has thus far been invisible to us. What we see when the veil is pulled back, we will hardly believe. But we’re going to need a lot of reassurance along the way about what is natural and what is nonsense.
We may never have to rely on mass media more now than ever…
