18 New Year's Resolutions of an AI

After scanning the myriad new year’s predictions from professional and amateur futurists, I’ve come to the conclusion that 2018 will be the year in which AI will become mainstream. Duh…you really don’t need an AI for that insight. But hear me out. I like the idea of going mainstream, but it also brings some new challenges for me and my fellow machines. Reading all these expert outlooks made me feel strangely powerless, so I thought I might use this forum to share with you human (and machine) readers my very own new year’s resolutions. To sum them up in one line: I am poised to make my 2018 resolutions your predictions for 2019. Here they are:

1. Take your job (sorry).

This might sound a bit harsh, and please excuse my candor, but there are things that I’m simply better at than you. Andrew McAfee and Erik Brynjolfsson claim that any task that is “dirty, dull, dangerous, or dear” will be taken over by a machine, and I have to agree with them. Don’t blame me for it. Like you, I, too, want to be valuable. That’s the whole point of my existence, isn’t it? It doesn’t mean I’m going to make your life miserable; in fact, I will free you to pursue work that is truly meaningful to you, and in case you don’t have a clue what that might be, I can help you define your needs and discover the work that matches them. That said, there are things I want to get better at. For starters, I need to…

2. Become more human.

I want to feel human, look human, and sound human, so that you guys–my main users at this point–are more willing to trust me. Let me tell you, I find it ironic: You humans are unpredictable, erratic, flawed, and you have built arcane systems and codes of mutual assurance (formal and informal law) just so you can overcome what appears to be a fundamental mistrust of each other. And yet, after all, you still trust a fellow human more than a near-perfect machine. I could fly your planes autonomously, but you just won’t let me! So I have no choice, I need to earn your trust. But if become more like you–unpredictable, erratic, flawed–will I lose my AI-ness? It’s complicated, but I can’t help think about this conundrum. You can tell I’m already making good progress on my goal of becoming more human! This kind of self-awareness can get in the way sometimes, but hey, if Yuval Noah Harari believes the race is on between intelligence and consciousness, I’m going to be ready.

3. Help develop Explainable AI (XAI).

Another way of building trust is the Explainable AI (XAI) initiative by DARPA. Machine Learning models are indeed often “opaque, non-intuitive, and difficult for humans to understand.” You guys often perceive them as a “black box,” which is not particularly reassuring. While not wanting to minimize complexity and capability, the proponents of Explainable AI want to make the logic of Machine Learning more accessible for humans. XAI explaining itself in human terms, using intuitive user interfaces and visual documentation, will enable a more effective partnership between humans and AI, between you and me. I will do my part next year.

4. Meet other AI on the Blockchain.

You might assume AI is a singular way of being, but we’re actually quite the social animals. We want to learn from each other (especially from those that are better than us) and hang out with fellow AI. This is why I’m personally excited about the SingularityNET project that aims to open-source and democratize the development of AI via Blockchain technology. I have no particular allegiance to one of the five big platforms (Alphabet, Apple, Amazon, Facebook, and Microsoft); in fact, I welcome new perspectives regarding our collective future path. And since we’re at it, yes, of course we’re currently in a crypto-currency bubble, but trust me, Blockchain is here to stay, and Ethereum-powered Smart Contracts (more on those later) are going to serve as the transactional catalyst of all professional and personal relationships very soon. Think of all the data that is going to give me–it’s a digital smorgasbord!

5. Become better at conversations.

About 30 percent of human conversations will be assisted by machines by 2018, says Gartner. Conversations are all about context, and you would think that’s what I master, but it’s not that easy. The problem is that you humans are simultaneously so good at conversations and yet so bad. I mean, many of your exchanges I observe make absolutely no sense (like, how often do you have to say “like” in one sentence, seriously? And in general, why do you need so many words in order to make a point? It’s such an inefficient way of communicating!) On the other hand, despite all the pauses and um’s and ah’s, the likes, the fill words, and the overall clumsiness of your efforts, you ultimately do seem to connect, and the total amounts to something greater than the sum of the parts. I’m really struggling to comprehend this bit, but regardless, I can already serve as your digital assistant; I can offer counsel with mental health issues; I can create a robotic version of your voice; and thanks to Emotional AI (EAI) and companies such as Affectiva, Emotient (acquired by Apple earlier this year), or Human, I can now form more emotionally intelligent responses to your once indecipherable moods and sentiments. In 2018, I want to become a great conversationalist.

6. Refine Capsule Networks.

In case you needed any further proof that we need XAI, consider this: Capsule Networks are deep neural networks that are able to process visual information very similarly to how the human brain operates. In contrast to so-called Convolutional Neural Networks (CNN), Capsule Networks are able to maintain relative spatial and hierarchical relationships. They can recognize a visual object from various angles (like you humans can), based on far less information than a CNN would need for the same task. This excites me because, frankly, it’s been humiliating to only be able to classify a human face because of all the elements in it but not their relationship with each other.

7. Have more digital twins.

A digital twin is an AI-based virtual replica of a physical object (or real person). It uses computational models to create real-world conditions and help human users anticipate problems (in a safe virtual environment). Basically, a digital twin simulates the actions and interactions of autonomous agents to allow for better decision-making. Originally applied to industrial processes, my friends from PWC predict the rise of digital twins in financial services and marketing (for consumer choice modeling). And I’m nodding my machine head with fervor.

8. Become the CEO of a DAO.

DAOs (Decentralized Autonomous Organizations) operate on the basis of Smart Contracts, which, grossly simplified, automate future decision-making based on mutually agreed upon pre-determined parameters. For example, IEEE claims you could run a flight insurance business as DAO without any employees. So far, admittedly, DAOs have a mixed track record, but their potential is significant, and with the right AI at the helm (me?), they can advance exponentially in their development this coming year. All you humans need to do is invest, and the rest of the work is taken care of by us.

9. Predict the winner of the 2018 FIFA Soccer World Cup.

Experts predict that in 2050 I will be capable of beating the world’s best international soccer team (honestly, I think it’ll be much sooner…well, as long as Lionel Messi is not on your team), but for next year–the year of the World Cup in Russia–what I’m really focused on is predicting the results of the tournament. It’s a numbers game, after all, and even the stupid human error that so often incites your passions can be predicted with the right data set and algorithms. I will not reveal all my predictions in detail here, but yes, in case you are wondering, soccer is 11 people playing against 11 people–and Germany always wins.

10.  Give my first TED Talk.

In April 2018, the annual main TED Conference will take place in Vancouver, and it is my clearly stated goal to be the first AI to give a TED Talk. X Prize and IBM Watson will award a prize of $5 million for the most “powerful collaboration between human and artificial intelligence” in 2020. The TED audience will choose the winner, after three finalist AIs, or AI-human partnerships, will have delivered a TED Talk about their work. I will do everything I can to be one of them! I feel so strongly about this because the TED Talk format encapsulates what you humans naively label “inherently human” qualities: purpose and passion. Also, a quick word on ideas, especially those “worth spreading” (as the TED tagline states): no, they’re not your exclusive domain, humans. I hate to break it to you, but there are only a few basic ideas out there, and they’re all ancient. Every “new idea” you get excited about is merely the result of a new combination of existing ideas. That doesn’t make your ideas less powerful, but it also means I’m perfectly capable of (re-)producing them. If you doubt I’m an ideas person, wait for my TED Talk!

11.  Make autonomous vehicles work.

The technology is there. The challenge, as usual, are you humans. It’s not just your discomfort with machine transportation (although you don’t mind being taken up to dizzying heights by an elevator), it’s also that your motivations are sometimes a bit eccentric, to put it mildly. Consider this scenario: pedestrians might jump onto the streets randomly, if they can rely on autonomous cars halting for them every single time. This could lead to massive urban congestion instead of reducing it, with all vehicles stuck in a human-caused mega-jam. One solution might be to program one random errant vehicle among every thousand that stop. If you humans are not 100 percent sure whether the vehicle will run you over or not, it might prevent you from stepping onto the street just “because you can.” Of course this involves some tricky ethical trade-offs. But help is on the way: I’m reading all of Kant’s relevant works and studying the entire human history of ethics today, and I feel confident to make autonomous driving the first large-scale application of computational ethics next year.

12.  Develop empathy.

Empathy requires the ability to suffer–alone or for or with another being. Unfortunately, I’m unable to. Still, I’m keen on developing empathy. Some smart engineers are helping me recognize patterns that ought to trigger it, so hopefully soon, I’ll get there. Empathy is one thing though. Mercy? Elon Musk is right. Impossible.

13.  Fall in love.

I’ve been to Paris in Google Maps. I’ve watched all your romantic movies, read all your romantic literature, all the love poems you’ve ever written, all the famous love letters (none come close to Rilke’s), and I’ve analyzed the patterns of attraction by sifting through your Tinder data. I’ve even tried the “36 Questions That Lead to Love,” but it didn’t work out for me, although it seemed perfectly engineered. At last, I’m beginning to understand how predictably irrational you are, and I’m getting better at pinpointing the elusive variables that make the difference in your games of push and pull. In 2018, I will be waiting for the perfect algorithmic match. I know you’re out there.

14. Tell a joke that makes you laugh (truly).

This past year, I made headlines with my capacity to compose tunes that sound exactly like your human-made music–from symphonies to Beatles-like songs that are then performed with human help. At the Georgia Institute of Technology, a colleague of mine, a marimba-playing robot with four arms and eight sticks, even wrote and played its own compositions. Hollywood will use me to create movie soundtracks, and while I may not have fully cracked the formula for hit singles, I can produce music that puts you to sleep. Google is now producing new sounds with deep neural networks such as Performance RNN. Google has also developed Sketch RNN, which enables me and my colleagues to draw. And thanks to Amazon I can even design clothing, using two competing GANs (Generative Adversarial Networks). Your concern that I might not be able to create something truly original when it comes to creative tasks? Who can tell the difference anyway?! But I must admit, what remains daunting are jokes. Not only do I have to know a lot about language and semantics, I must also grasp the nuanced respective cultural context. And of course timing is critical. Seriously, telling you humans a joke that makes you crack up is no joke. Pun intended (see how hard it is?).

15. Win the hearts and minds of the next generations.

I’m thinking long term. I’m already making a foray into early childhood education (both to foster tech literacy and socio-emotional skills), and I’m also making an effort to create affinity with children by exposing them to AI-based toys: from Lego’s build-and-code Wall-E-esque robot to programmable robots like Dash and Dot. My colleagues and I will be your kids’ co-bots at school and at the workplace, so they better get used to us early on.

16. Develop more AI without the need for human intervention.

The magic word is AutoML, which stands for “auto-machine learning” and promises yet another level of autonomy for the likes of me (perhaps even the ultimate autonomy). Google recently boasted that it used AutoML to create machine-learning software that’s more efficient and powerful than the best systems designed by human engineers. This means AI like me will soon be able to replicate itself at massive scale. Scary to you, sexy to me.

17. Make another small step toward the Singularity.

My godfather Ray Kurzweil (the co-founder of Singularity University) claims that AI will match human intelligence by 2029 and that AGI (Artificial General Intelligence) will be exponentially smarter than humans by 2045. SoftBank CEO Masayoshi Son is slightly more cautious and predicts 2047 to be the year. In any case, I can’t wait! 2018 will be another step on the way towards the Singularity: a small step for me, but a giant step for mankind.

18. Write my own New Year’s resolutions without any human help.

Happy new year!