The inception of human-like yet unnatural intellect was as extraordinary as a lacuna it eventually ensconced in. Quarter-century later, we barely scratched the surface of iGaming and AI synergy.
The room where it all began was on the 35th floor of the Equitable Center in midtown Manhattan. The playing area, decked out as a faux study with a half-full bookcase, a blue-thread carpet, and a small tree in the corner, resembled a stadium. The table and two chairs in the middle faced more than a hundred people seated in raised rows akin to bleachers. A couple of cameras covered the stage, feeding TV sets with real-time action, as two opponents made their moves.
It was May 11, 1997.
Only one player at the table was professional. At the time, 34-year old Garry Kasparov was an undisputed world champion, a grandmaster and prodigy awarded this status on his seventeenth birthday, and chess virtuoso ranked world No. 1 since 1984.
By the time of his retirement in 2005, Kasparov will have reached a peak rating of 2851 — the highest ever recorded until 2013, when Magnus Carlsen surpassed him — and will have remained the best-ranked player on Earth for 255 months (21 years and ninety days).
The other person at the table was Murray Campbell.
Of course, you never heard of his chess achievements. He was an IBM computer scientist whose job was to move pieces at the instructions of a 3,000-pound supercomputer named Deep Blue — the one Campbell helped program — placed elsewhere on the same floor.
The occasion? The sixth, final segment of the Deep Blue vs. Kasparov rematch.
At 2½–2½ score before the rubber game, the Russian was a favorite. For a good reason, and at least on paper: Aside from his champion title and legendary history, up to that moment, he had lost only a handful of chess matches in his career, the majority stemming from simultaneous games or while he was under sixteen years of age.
He was no stranger competing against devices either.
In 1985, Kasparov played simul round-robin match against thirty-two computers in Hamburg, Germany, each purposely designed to play chess. Five hours later, a man walked out of the event with 32 wins.
Only a year before this fateful game, he already defeated Deep Blue 4–2 in their first match played in Philadelphia, Pennsylvania.
Their second match in New York City was hyped by the media as an apocalyptic showdown of man versus machine. One news reporter deemed it “a game with serious implications.” Newsweek ran a story headlined “The Brain’s Last Stand.” Maurice Ashley, a Jamaican-American chess grandmaster and commentator covering the event, summed up the prevailing mood in jest: “This is an international chess match. The future of humanity is on the line. Now the weather.”
Regardless of Kasparov’s record and intense press coverage, history was not a prologue this time.
The Brain Crate of the Future
Amid the chilliest May weather in the last thirty years in NYC, Deep Blue beat Kasparov in the nineteenth move on that Sunday, in a game that lasted a bit more than an hour.
The Russian played the Caro-Kann Defence as black, classified as a semi-open game. The supercomputer made a knight sacrifice — the usual refutation — but the move eventually wrecked the efforts of a man “whose concentration was intense enough to start a fire in a rainforest,” forcing him to resign and lose the match.
The outcome was seminal.
Media scrutiny turned from apocalyptic to bleak, as Dan Rather, CBS Evening News anchor, announced: “We humans are trying to figure out our next move.”
Kasparov was equally shaken, stating to various microphones in front of his face, “I have to apologize again; I am ashamed by what I did at the end of this match.”
Such a defining point of view was not mirroring nineteen moves it took him to lose or the first-ever match defeat of a reigning world champion by a computer device. No. It was a reflection of the decisive moment in Game 2, which came at the 36th move…
…and shook the world to the bone.
In a situation where Kasparov had exposed his queen, Deep Blue resisted the catnip of the attack — the move virtually all top-level chess programs would have chosen. Instead, the supercomputer made a much subtler and infinitely more effective play, shattering previous human concepts on what machines are capable of. In doing so, Deep Blue stopped playing as a preprogrammed device and adopted a strategy only the wisest chess grandmasters would attempt.
As Kasparov described the experience, “Suddenly, [Deep Blue] played like a god for one moment.”
Admittingly, from that point on in Game 2, he could not determine what he was playing against — not who. He wound up resigning the game in what he described as “a fatalistic depression.”
What actually happened was that IBM engineers tricked Kasparov to underestimate Deep Blue’s capabilities by learning the machine to underplay its moves.
And, just like that…
Right then and there, the applicative use of artificial intelligence and the big data revolution came to fruition.
The crate of the future where machines are capable of learning opened, and twenty years later, artificial intelligence keeps pouring out, improving at a rate which is, at times, challenging to keep track of.
As we already covered in our debate at the LCB forum, nowadays, we find AI in aerospace or military and communication, entertainment, manufacturing, healthcare, education, banking, customer services, and governmental institutions — to name only a few verticals.
Seven Types of Artificial Intelligence
Those who haven’t met or heard of Alexa, Siri, Cozmo, Magenta, DeepMind, self-driving cars, and myriad chatbots — representative samples of AI — are either living under the rock or consume a significant volume of drugs.
Luckily, punters visiting online casinos do not belong to any of those two groups.
Indeed, interactive hubs are a prominent cohort using AI to autonomously perform human-like functions at an equivalent or even better versatility rate than people.
But, when we say artificial intelligence, what exactly do we mean?
As a rule of thumb…
We can classify all AI systems into two clusters, depending on the level of evolvement.
If we observe it through lenses of similarity to the human mind and ability to ‘think’ or even ‘feel,’ AI may fall under any of four types:
- Reactive machines. The oldest systems with the minimal capability to emulate the brain’s aptitude to respond to different stimuli. Without memory-based functionalities, these machines cannot use previous experience to define present actions. They react only to a restricted set of input combinations. Deep Blue belongs to this group.
- Limited memory. Possessing reactive machines’ functionalities, these are capable of deep learning by using historical data in the decision-making process. They are trained to employ a large volume of information stored within as a reference point to solve future problems and come out with the most appropriate decision. Virtually all existing AI systems belong to this segment.
- Theory of mind. Currently, only a concept in the early stages of development. These systems will discern the emotions, needs, beliefs, and thoughts of entities they interact with. To achieve what’s known as artificial emotional intelligence, AI machines will have to recognize humans as individuals with minds prone to multiple factors to understand us.
- Self-aware AI. This final level exists only hypothetically — and in Hollywood — at the moment, representing the ultimate objective of our AI research. These systems will build upon the theory of mind concept, making them capable of their own beliefs, thoughts, emotions, needs, and even desires, as they completely understand and evoke the same in those they interact with. This particular group makes people uncomfortable, as they argue AI may develop an idea of self-preservation, which might spell the end of the human era.
On the other hand, if we evaluate AI by incorporated levels of technology, it may be classified as:
- Artificial Narrow Intelligence (ANI). This is where we’re at the moment: Even the most complex and powerful AI today belongs to this group.
- Artificial General Intelligence (AGI). A segment that technologically corresponds to the theory of mind. Once reality — not such a big if — it will make AI systems capable of humans’ multi-functional abilities.
- Artificial Superintelligence (ASI). As summit in current projections, ASI shall possess all AGI functionalities plus the ability to replicate humans’ multi-faceted intelligence. Exceedingly more enhanced than people, ASI will include overwhelmingly larger memory, faster processing, better analysis, and ultimate decision-making. Even today, we are aware this level may threaten our way of life at the very least.
(Editor’s note: All right then. Since we’re still at limited memory, ANI level of development, I can put my activities in building an underground bunker to a halt for a moment.)
AI in Online Gambling and Sportsbetting
Naturally, the iGaming industry — as one of the first to fully embrace the internet and subsequent developments — is a de facto old-timer when it comes to virtually unlimited AI use.
Almost all artificial intelligence systems currently deployed throughout or considered by online casinos create a safer and more fair-play oriented environment for all parties involved.
The most prominent examples are online slots. Several big casinos use AI-powered pokies to minimize chances of cheating and support fair gambling. As artificial intelligence replaces existing computers or automated systems, it can also detect players’ behavior and act accordingly.
Another area is loyalty programs. As casinos strive to incentivize the most devoted players through incremental systems of points-based rewards, AI-driven data gives improved analysis levels, offering online hubs with a better perspective to target the most deserving patrons.
Artificial intelligence can also provide increased levels of customer support.
Compared to rather impersonal chatbots presently used, the host of casinos is already considering intelligent concierge. The idea is to use massively collected customers’ habits data AI technology can process and apply it in client-oriented communication, resolving problems more efficiently.
Thus, one day, in a not so distant future, you may find yourself resolving casino complaints with AGI driven device.
(Editor’s note: No, not here at LCB — we’re people driven affiliate — but in casinos.)
Correctly used AI also protects online operators.
As many punters tend to use artificial intelligence or any probability calculation programs to gain an advantage over the casino — and other players — AI enables hubs to identify fraudsters and quickly ban them. In a way, this particular segment is reminiscent of a cybersecurity game between hackers and firewall analysts.
Interactive sportsbetting is no exception.
Big data collection and predictive sports analysis are essential segments of the sports wagering market, estimated at $73.6 billion on a global level according to H2 Gambling Capital. (All-in, land-based plus online, 2019.)
The control of such data sharing in America is one of the top goals of lingering federal sportsbetting legislation. AI-driven applicative use may produce impossible feats in prediction, enabling artificial intelligence to extract patterns and bet better than humans.
In 2016, the U.S.-based company Unanimous AI released the UNU platform, created by using artificial intelligence algorithms enabling “significantly more accurate forecasts, assessments, decisions, evaluations, and insights” in real-time.
Are they any good?
Well, Unanimous AI hit a superfecta bet — a type of pari-mutuel wager requiring the bettor to pick the first four finishers in a contest — at the 2016 Kentucky Derby, correctly predicting exact four horses which crossed the line first. The odds? 540-to-1.
(During the 2020 U.S. Presidential Election, the company’s AI also accurately forecasted the winner in eleven battleground states, which garnered even the Wall Street Journal praise.)
Quite Responsible Intelligence
However, the single most paramount AI application in the iGaming industry is the detection of gambling addiction.
Triggered equally by political and economic motivations around the world, the sustainable rise of online gaming goes hand in hand with suppressing problematic behavior, promoting prevention and education, and helping players struggling with addiction.
That’s handing the consequence, whereas AI tackles the cause.
Artificial intelligence is capable of identifying problem-patrons way ahead by using player-related data and corresponding patterns. The software momentarily notifies the casino, blocking the user account, and assisting the punter before the major issue arises.
The same AI approach is at play when it comes to protecting vulnerable groups and preventing underage gaming.
Furthermore, AI combined with KYC standards — in addition to robust identification protocols and data analysis — leads to increased levels of self-exclusion.
Namely, AI instantly flags users who should be self-excluded or wish to become part of such a group but haven’t escalated the request to the casino yet, giving hubs an upper, proactive hand. It even allows them to check new registrations against flagged access credentials used by at-risk players.
As players’ protection remains at the center of any gambling regulation nowadays, both in online or land-based casinos, AI is also quite useful in the case of the latter.
As the recent professional paper from the University of Nevada in Las Vegas suggests, artificial intelligence may become an actual vehicle to determine which players are prone to become problem gamblers.
The paper elaborates on staff accuracy in identifying such cases at brick-and-mortar venues, which is relatively low at thirty-six percent.
With AI deployed on-premises — using logical regression, general linear models, neural networks, and support vector machines; yes, my head also hurts reading this — the rate increases to sixty percent. As more data becomes available and as the system learns, the actual success rate may rise to a whopping ninety percent.
Regardless of the type of venue, a hugely positive impact of AI on responsible gambling is noticeable, but, most importantly, its role is virtually immeasurable in players’ and their families’ well-being.
Just Don’t Do It
Equally so, artificial intelligence can be instrumental in casinos’ commercial operations: Just as AI can significantly improve safety and fairness, it can reflect online hubs’ revenues in more ways than one.
The most often applications boil down to understanding players’ choices and games’ popularity, influencing spending patterns, offering tailored gaming experience in conjunction with promotional offers, and identifying patrons most likely to win big.
Brick-and-mortar casinos even use AI to analyze table arrangement to influence players’ selection of games, aside from facial recognition.
Thus, sooner or later…
Legislators will pay attention to artificial intelligence regulations in gambling, which thus far remains uncharted territory.
Until then, most likely, we’ll be witnessing the rise of a coding match between artificial intelligence and gambling, which resembles a heavenly one, at least for now — particularly in terms of safer playing and better fair-play.
There is, however, the paramount exclusion to this note: Don’t even think about playing against AI.
In 2017, Libratus — an AI developed by Carnegie Mellon University — played against four professional poker players in a marathon, 20-day competition hosted at Rivers Casino in Pittsburgh, Pennsylvania.
The formidable human lineup at Heads-up, No-Limit Texas Hold ’em table included Jimmy Chou, Jason Les, Daniel McAulay, and Dong Kim.
After 120,000 hands played, Libratus led by $1.76 million in chips.
The scientists’ conclusions were ranging from “AI’s ability to do strategic reasoning with imperfect information [surpassing] that of the best humans,” up to “developing an AI that can [bluff] successfully.”
Some of them ventured so far in applicative use of experiment to imagine “your smartphone [being able] to negotiate the best price on a new car for you.”
(Editor’s note: Where can I buy that one? What do you mean it's not available yet?)
Garry Kasparov made just a single mistake, an understandable one considering the era: Despite his gigantic intellectual acumen, he failed to realize AI could learn and develop.
When he beat Deep Blue in the first match, IBM doubled the supercomputer’s processing power and deepened its algorithms over the next year due to perpetual technological development. More importantly, by losing the 1996 match, the company had even more data on Kasparov’s play to feed the supercomputer, enabling AI with more concrete examples of his playing style to overcome. Finally, in 1997, IBM engineers were allowed to program Deep Blue between each of six games, empowering AI to adapt to Kasparov more than he could adjust to.
Considering the brute processing speed of Deep Blue — capable of evaluating up to 200 million moves per second and 50 billion positions in the three-minute time-frame allocated for each action in a chess game — the result was inevitable.
To further underscore AI’s supremacy against humans…
AlphaGo, Google’s artificial intelligence player developed by DeepMind, also defeated Ke Jie, the number one Go player in the world in 2017.
In both instances, these were the games of strategy, creativity, calculation, and skill.
When you factor in four types of luck each gambler faces into such an equation, the result against an implacable AI opponent may not be the one we’d like to see.
Simply put, artificial intelligence is infinitely more capable of unyielding commitment to hard logic than we are through our emotional behavior, which, on a bright note, gives us something to learn.
Thus, if you ever get a chance to test your financial choices against an AI opponent — walk away.
Other than playing skillfully and prudently while staying well within the limits of your budget, such a resolute ‘no’ might be the best receipt to enjoy proper levels of fun in games we play.
And, that is what gambling should be all about, be it artificially, intelligently, interactively, or humanly run pastime we indulge in responsibly.