Today, Siri, Alexa, and Google Assistant are sometimes useful, sometimes fun, but honestly, very limited.
They often return a generic answer or a bland web search result when we ask them questions. Considering the massive resources behind Apple, Amazon, and Google, it's surprising how little progress these tools have made.
COVID-19 pandemic
During the COVID-19 pandemic, many of us spent months isolated from society. The old times when we used to retreat into the mountains for a few days, sit around a fire, talk, eat, drink, play the guitar, and sing under the stars might feel as distant to today's teenager as the nearest galaxy.
The COVID epidemic will leave a deep mark, and that shift will influence how the next generation relates to technology.
Gathered Data
I’ve bought countless items on Amazon, so they know my interests. I’ve read Kindle books and listened to hundreds of Audible titles.
I searched Google thousands of times and browsed academic papers through Google Scholar.
Given all that data, I could have endless and engaging conversations with someone or something that matches my interests. I need that kind of companionship.
Yet Alexa, Siri, and Google Assistant ignore this need entirely.
Short and Long Term Memory
They lack basic short-term memory, cannot continue a conversation, and can’t connect today’s question with what I asked this morning or yesterday. They cannot tell a relevant story based on what I’ve read, written, studied in college, or even who my friends and professional contacts are.
They don’t connect people based on deep interests or alignments. They don’t recommend conversations, articles, or human networks. They don’t even distinguish me from one of my kids. The other day, I had to unsubscribe from three different “fart skills” on Alexa.
My Ideal Personal Assistant
I want a digital assistant trained not only on a general corpus of books, music, films, and research, but also specifically on everything I’ve read and written, on the total knowledge I’ve consumed.
It should respond to a unique name in a clear voice, with the accent and intonation of my choice.
It should recognize my face, sense my emotional state, and adjust tone and engagement accordingly.
If someone else is speaking, it should know that and invoke their personal assistant, protecting privacy and individual context.
It should have a unique face generated just for me via the generative adversarial network (GAN) or something similar, so I can quickly recognize it and intuitively read its mood or expression.
It should also be able to reach millions of others on my behalf, making introductions based on what both parties truly need, factoring in timing, language, and culture.
I want to start my day with AIKO (child of AI in Japanese), which gives me a personalized news summary, including stock updates and curated recommendations. I want to pause, resume, ask questions, skip what I don't care about, and dig deeper when interested. I want to receive a daily article or book suggestion tied to my current interest. If someone out there is aligned with me, the assistant should arrange a message or a meeting.
A sense of humor and tact would be essential. All this should happen before I sit down at my desk.
Will Apple, Google, or Amazon build this? Maybe. Eventually.
But the opportunity is open to others. Just as each person is unique, so are their digital needs. What I describe may not appeal to all, but it will resonate with some. Enough to matter.
Things to pay attention to
- - OpenAI's GPT-3 based on the Transformer model
- - retrieval-based memory systems (RAG)
- - Vector embedding of documents
- - iOS with Core ML, Apple Neural Engine
- - DALL·E
- - generative adversarial network (GAN)
I'd love to connect if you're working on something like this.
