While AI is continuing to innovate and disrupt industries from construction to gene therapy, impediments plague the current adoption of AI through to mass application. AI originally developed for the board game GO and the Video game DOTA 2 are making huge waves in not only research but real world application.
If you are unfamiliar with both AlphaGo or OpenAI please see these articles that explain all about them:
My goal in this article is to explain three major pitfalls in AI, and how the advancements in game-playing AI will help to address these issues, allowing AI to be used much more widely and effectively in all industries.
The first of these issues is the problem that the majority of AI needs an incredibly large amount of data in order to be able to complete the tasks assigned. Currently, most of the major uses of AI are in areas where vast amount of data can be collected and fed into the AI.
These industries include image recognition, and search preference creation/suggestion. Neil Lawrence, a Professor of Machine Learning at the University of Sheffield and part of Amazon’s AI team echoes these sentiments, stating “if you look at all the applications domains where deep learning is successful you’ll see they’re domains where we can acquire a lot of data,". (Vincent James).
This is an area where both the advancements in AI made by OpenAI and AlphaGo are remarkable. Both OpenAI and AlphaGO Zero were designed without any data, but rather given the parameters of the games, and learning took place via playing altered versions of themselves. This could open countless doors for AI to impact industries that currently lack the vast amounts of data currently needed. Alternatively, revolutionary changes could occur in industries like image recognition where the data can be collected, but the speed of making decisions without waiting for it all to be collected would lead to more powerful and efficient results.
The second issue was eloquently stated by Rodney A. Brooks (Professor of Robotics (emeritus) at MIT) at the Thirteenth National Conference on Artificial Intelligence in 1996:
“Biological systems can adapt to new environments--not perfectly, they die in some environments, but often they can adapt. Currently our programs are very brittle, and certainly a program compiled for one architecture cannot run on another architecture.”
Although this issue was raised in 1996, it is still one of the biggest issues facing AI today. AI research scientist with Google, Raja Hadsell explains, “There is no neural network in the world, and no method right now that can be trained to identify objects and images, play space invaders and listen to music.”. (Vincent James).
Although both AlphaGo and OpenAI were designed for Go and DOTA 2 respectively, OpenAI can accomplish many tasks within the game such as purchasing items and using abilities. For example, AlphaGo Zero was able to best the world's top Chess AI, StockFish, after only four hours of training. (Gibbs Samuel). Both Google and OpenAI are companies with their hands in many other projects that either focus solely on AI or could benefit hugely from AI. Thus, in the production of AlphaGo and OpenAI, both companies kept the idea of adaptability in mind.
The third and final issue is the idea that AI is currently a ‘black box’. Most people aren’t able to understand how advanced AI come to the conclusions that are made. As Will Knight put it in the MIT Technology review, “No one really knows how the most advanced algorithms do what they do. That could be a problem.” (Will Knight 54). Although it is understood the parameters and rules put in play, it is still exceptionally hard to see how the AI is actually “thinking” and coming to its conclusions.
This issue causes problems twofold. Firstly, it is hard for people to work collaboratively with the AI if the person is unable to understand how the AI came to its conclusions. Secondly, this ‘black box’ issue has fostered and maintained public distrust towards AI, and thus widespread adoption is much harder to attain. Due to the easy to understand grasp and nature of both Go and DOTA 2 as a board game and videogame respectively it allows the public to digest AI advancements in a easier way.
Additionally one of the aspects of OpenAI’s AI is that although it does not explain why it executes the moves it does, top tier players are able to understand its rationale. This could potentially lead to the idea of gamifying AI and its tasks in order for high level human collaborators to better understand and work alongside the AI.
Before continuing into the predictions for game playing AI’s impacts in advertising and business development, it is important to preface any predictions with the understanding that the AI industry is constantly changing and even the foremost experts such as Bostrom can be grossly inaccurate with their predictions.
The Seven Deadly Sins of AI Predictions written by Rodney Brooks covers many of these pitfalls to avoid, and in examining the next sections it is best to keep them in mind. Below I have outlined three of the more apt seven ‘sins’ and given very brief summaries of each.
1. Performance versus competence
Separating the fact that, unlike the human ability to do a task and all the other tasks that would usually accompany that ability, computers cannot.
2. Imagining magic
The idea that because we do not know the limitations of something, we imagine it to be limitless.
How OpenAI and AlphaGo can affect jobs:
Due to the nature of both DOTA 2 and GO, AI advancements and victories against humans represent much more than simply the strength of large amounts of computation. Rather, what is required is understanding nuances, taking calculated risks, and adjusting to changing conditions not previously expected. Thus, the next part of my paper will be to find out just how likely it is that projects like AlphaGo and OpenAI will have an impact on jobs in advertising and business development. Examining how and whether game-playing AIs really do point to the likelihood of AIs that can operate in the different contexts of business, and advertising will allow me to accurately argue a further point; that this new AI revolution will allow us to accomplish what we thought was previously not possible. Before exploring the predictions of how AlphaGo and OpenAI can impact business development and advertising, it is important to not “imagine magic” or confuse “performance with competence”. Although many of their skills are beneficial for multiple industries and may in the future be used to disrupt many fields, currently AlphaGO is only able to operate in Go and OpenAI only in DOTA 2.
Politics, Betrayal, Power - what do these words have to do with AI? Well, House of Cards, Netflix’s first big foray into original content production was created in part by an AI analyzing the viewing habits of 33 million users, and using such feedback to custom create and gauge the success that House of Cards would have.
Although this type of big data collection and utilization isn’t anything new, with the increasing fact that data giants like Google, Amazon, and Netflix are starting to create their own content, they could end up “heavily shaping creative decisions of directors and writers as well.” (Leber Jessica).
Exploring the potential impacts AlphaGo can have in business development, it is able to both compute an extreme amount of data, coming up with different moves and strategies whilst also giving the percent success rate of each option.
This could be extremely useful in something like Netflix’s House of Cards scenario. For example, if it was determined that viewers would love a new crime show concept, but they couldn’t decide on who would star, they could input the different actors/actresses into an AI like AlphaGo and see how the likely success would change.
Additionally, due to the fact that AlphaGo is extremely fast in coming up with different ’moves’, and is able to go through the success of the different potential outcomes each ‘move’ would have, it could greatly help companies when they are met with crisis. For large companies, when crisis hits, each decision could have catastrophic outcomes if handled incorrectly.
Having a tool like AlphaGo which is transparent in its decisions eliminating the ‘black box’ of AI and creating many plans to success, would allow executives to test many different decisions and thus help guide them to the most beneficial outcome and plan.
Building off of another area within business development, AlphaGo could be very impactful is helping to decide whether to expand into new areas or create new products. The upgraded version of AlphaGo, AlphaGo Zero, as mentioned in the background section, was recently created without any training data and defeated AlphaGo 100-0.
This ability to create successful strategies and outcomes, without training data, can be hugely beneficial for businesses looking to expand into new industries or develop new products where small amounts of data are had. This could not only lead to more profitable expansions for companies, but also push what a business previously thought it could be capable of.
To reiterate what was said above about AlphaGo Zero, “By not using human data or human expertise, we’ve actually removed the constraints of human knowledge,” “It’s able to create knowledge for itself from first principles.” (Silver).
If AlphaGo Zero was given enough time, it would be able to come up with complex plans suiting many different scenarios without ever needing any data.
Adaptability and innovation are not only vital skills needed to be successful in life but also are skills which are most valued in Advertising.
To be successful in advertising, there are many things to consider, especially in our current times where trends change rapidly, political climates are constantly shifting, and influence can be ephemeral. According to AdWeek, “U.S. adults spent 10 hours, 39 minutes a day consuming media in the first quarter of 2016.”(Lynch Jason). All that time adds up to more and more consumption of advertisements.
Therefore, being able to create something new and fresh is vital to the success of an advertisement.
OpenAI’s ability to create a successful AI in DOTA 2 without all the knowledge of its target, and a constantly changing environment where each move’s success can be changed as each second passes, lends itself very well to advertisement. Dynamic Ad Insertion, the ability to change the content of an ad dynamically based on the reception of its’ target audience is the future of advertising (Yan Melody).
Creating advertisements that can morph into better versions of themselves based on huge amounts of data will allow those advertisements to hit target audiences sooner, or stay relevant in our constantly changing landscape.
OpenAI lends itself to being able to help create this future. Instead of DOTA 2 opponent moves and trends, it could take in Twitter, Facebook and Instagram reactions to an ad, and then dynamically change it so that in its next viewing cycle it would be improved for the audience.
The fact that OpenAI achieved such dominance without training sets means that it can come to these very human and non-human strategies completely on its own. This can especially impact the landscape of advertisements by creating unique content which has human characteristics but pushes the envelope further, expanding the knowledge base of what a successful ad will look like.
This will create a cycle of further growth as these new ads are created, analyzed and thus used to create even newer and better ads.
In addition, and potentially a matter of trepidation, OpenAI is always making each move with the constant goal to win, taking in the goals of the opposing players and using them to exploit their vulnerabilities.
This was demonstrated when OpenAI willingly had itself damaged by Dendi in the game, in order to bait him into following him closer to its base where it turned on him with upgraded abilities and killed him swiftly.
These advancements of OpenAI to analyze more than just the data has the potential to influence future AI specifically created for advertising to analyze the deeper goals and exploits of the consumer.
Thus we could be creating a future where we are being exposed to changing advertisements, that do not only target our demographics, but also our deepest emotions, strengths and vulnerabilities.
About The Athor:
Aanikh is currently the Co-Founder and COO of Trufan, a social intelligence platform that helps brands or influencers manage their key followers, sell directly to top fans and identify new audiences to target. At the age of 13, Aanikh created his first business and social venture - UndrTheRadr (UTR). At age 15, Aanikh became the youngest social entrepreneur ever to receive venture capital funding on CBC’s Dragons’ Den. UTR reached the #2 most downloaded App on the iTunes App Store beating out Angry Birds and Minecraft. At age 16 he was named and awarded Canada's Top 20 under 20 and is currently on leave from Stanford University where we was studying Management Sciences and Engineering.