The AI Chatbot Wars: Who Will Rule the Iron Throne of the AI Revolution?
If you are reading this article, you probably recently read, used or just even heard about ChatGPT, which after its official launch (November 30, 2022) it is a tool/platform which has become viral and made an all time record for the time it took to reach 1 million users. It took just 5 days! As a comparison with other technology platform disruptors and how long those took to reach 1 million users (source: statista.com):
If you also take a look at the recent search behavior globally, the search for the keyword “chatgpt” has exploded even when you compare it with a very popular keyword such as travel (source: Google Trends):
?
Why has ChatGPT created such buzz, even though Artificial Intelligence has been around, in some form at least, for quite some time now?
If you belong to the ones who have used ChatGPT, you have probably been amazed by its abilities with regards to what it can do as a conversational digital assistant. Abilities to understand a request, form a dialogue and produce relevant, tangible and (mostly, but not always though) accurate, results. Something that no other solution open to the public has done in the past to this extent and also by using a conversational interface. You have also probably and immediately thought of ethical, philosophical, as well as practical, legal and regulatory questions which have or will arise in the future by having such Artificial Intelligence (AI) tools available.
You have also probably wondered how such tools/solutions will impact the internet, internet Search or as I like to call it, “information and machine interaction” and content creation as we know it today and more specifically how the #1 Search company in the world, namely Google, will respond to this technology disruption.
There have been steep developments occurring in the past few months. In January of 2023, Microsoft became one of the main collaborators and investors to OpenAI, the company behind ChatGPT as well as other AI tools.platforms such as Dall-E2 and soon thereafter, on February 6, 2023 Google’s CEO himself introduced “Bard” as Google’s response to ChatGPT.
Before though we deep dive further in all of the above and see also if all the recent hype behind ChatGPT is warranted, it is very useful to take a trip down memory lane and see the evolution of the internet and internet search, as I also experienced it first hand during my career journey thus far.
The evolution of the internet and internet search through my eyes and my journey
My first interaction with the internet was back in the 90’s, during my National Technical University of Athens (NTUA) days when I was studying for my first degree in Electrical Engineering. I remember it like it was yesterday when we were experimenting with a few fellow students of mine at a Computer Lab using our UNIX workstations and accessing files on an MIT server via a browser called Mosaic and later on Netscape at that time. That right there was my first “wow” moment, when I realized – even if Ι have to admit it was just slightly that I realized it - how computer and server interconnectivity could potentially revolutionize the access to information.
It was also thereafter, while I was studying for my Electrical Engineering Master’s degree at Stanford University, that I witnessed the progression of internet information retrieval moving from web directories, like the Yahoo directory and DMOZ, to search engines. In the beginning, search engines were not very accurate in responding to search queries. The most decent one back then and most popular because of that, was “Alta Vista” which was later on acquired by Yahoo. And then, something disruptive came along within Stanford. There was word of mouth about a search engine which was developed by two students who graduated same year as me from but from the neighboring Computer Science department, namely Sergey Brin and Larry Page. That search engine had a funny, yet catchy name. Google... And it was indeed a search engine which was the first one where you saw it bringing relevant and thus quality results upon your query, based on its patented PageRank (named after Larry Page) algorithm!
Afterwards, Google started developing other products as well but, at least in the beginning, it seemed that there was no concrete strategy behind those efforts, nor a monetization model, but rather a more geeky/tech approach in offering digital and disruptive solutions on the internet. In those old days, Google used to release their products in beta version for an extended period of time in order to gain user base, have essentially a form of crowd testing and also collect valuable data from its users. I still remember the time when I registered for my Gmail account back when you had to receive an invitation from an existing user and that product was on beta for a very long time. And then, slowly but steadily, Google started putting the pieces together, developing its internet ads products and therefore its main monetization model, developed an OS for mobile devices (Android) and in general made more evident its overall strategy to become the #1 search and overall #1 internet company in the world, having also collected an extremely large number of data around its internet users and their behaviors, through all of its suite of tools/platforms such as YouTube, Google Maps, Google Photos, Gmail and others.
Since then, I witnessed Google’s main product (the search engine) developing into a lot more than a search engine. Fast forward to today, Google’s Search Engine market share worldwide exceeds 92% - and for mobile it even surpasses 96% (Statista figures dated January 2023), whereas 60% of its overall revenue comes from its Ads business (not only from search but from all other products which have incorporated its Ads framework):
Artificial Intelligence is nothing new to Google. It has been using it in order to optimize its search engine and all of its products, so as to optimize in turn the value being offered to its users. It has also developed products such as the Google Assistant which utilizes a conversational HCI (Human Computer Interaction). Such HCI is also nothing new to the user. Apart from the Google Assistance, there are also other virtual assistants as commercial products such as Siri from Apple and Alexa from Amazon.
Voice Search was one of the next big things even back in 2018 and one of the digital marketing trends which I had mentioned during my imc.2018 (2018 iab Hellas international marketing conference) keynote speech. What is indeed though new and impressive with ChatGPT is its form of intelligence in order to understand each query, synthesize information and present its response, many times without even repeating itself.
Artificial Intelligence types and what is now different with ChatGPT
Before we move further though, it is worth taking a look at the main categorization for AI types/stages which is the following:
So far, AI has had great applications at ANI, where the machine is trained in order to tackle task specific problems and is actually very good at it. Such examples include reading medical images and performing an initial diagnosis/assessment, understanding single requests (Siri, Alexa, Google Assistant and other chatbots such as Meta’s Messenger) and responding accordingly, performing image search, OCR, etc.
Note also that one such AI application was part of an actual commercial product by Apple back in 1996, the year when I started my professional career working for Apple. That product was Apple’s “Newton” which was a PDA (Personal Digital Assistant) Apple had created and was way ahead of its time and much better than popular PDAs of that time such as the Palm Pilot. Newton’s main differentiation from other PDAs was the fact that although other PDAs were expecting the letters to be written following a specific pattern which was instructed by the manufacturer, Newton was accepting each user’s own handwriting and was gradually learning it in order to best understand and convert it to the respective computer fonts. Apple’s Newton was discontinued by Steve Jobs one year upon his return to Apple in 1998.
What is though indeed really disruptive with ChatGPT is that, for the first time, such a platform released for wide use, moves, even slightly, towards AGI and is able to hold as much of a coherent conversation as possible with a human being and use its knowledge (currently it has been trained until end of 2021) in order to respond to requests (not always accurately but still amazes you with the results it produces most of the times). ChatGPT utilizes the information and knowledge for which it has been trained in order to combine it and create an appropriate relevant response to the query. Does this mean we have already mastered AGI with ChatGPT? Hardly! We have simply started scratching its surface. If you notice ChatGPT’s responses, in most times it simply retrieves information and knowledge, assembles them and responds with a rather “diplomatic” response. And many times, as you will see further down, ChatGPT’s responses are inaccurate, which raises trust, as well as security issues. The most important thing though to point out here is that we are now at least starting to see machines moving towards the AGI space.
What is ChatGPT?
Chat Generative Pre-Trained Transformer, commonly called ChatGPT, is a chatbot launched by OpenAI (the inventor of other popular AI tools as well such as DALL-E and DALL-E2 which generates digital images from a natural language input) in November 2022. It is built on top of OpenAI's GPT-3 family of large language models and is fine-tuned with both supervised and reinforcement learning techniques.
It is claimed that ChatGPT has been trained on 175 billion parameters, 300 billion words, and 570 GB of data. This is massive amount of Big Data which has been used to perform its supervised and reinforced learning.
It is essentially an extremely versatile chatbot that is designed to make it easy for the user to perform a variety of tasks, such as:
Write an essay, blog post, novel, opinion, email, SoMe posts, pitches, etc
ChatGPT is very easy to use due to its conversational HCI (Human Computer Interface). You simply type in what you want, engage in a conversation and ChatGPT takes care of the rest. This HCI is very easy to use because it responds to a basic human reaction which is a conversational communication, which is also in my view one main reason why ChatGPT has become so viral. At the same time though, exactly as if you are interacting and communicating with another human being, the answer you are getting is 1 answer per query which means that there is a big “trust element” between the parties of that conversation. If you ask someone something, your brain immediately assesses the validity of the response based on the trust you have towards your conversation participant. This is one main challenge, along with ethical and other dimensions, which arises through such HCI. Google’s internet search works a bit differently. Google brings you a bunch of relevant results per query stating the exact source/publisher behind each answer and lets the user judge and assess which information to trust and thus use.
ChatGPT isn’t connected to the internet and does not have access to external information. Instead, it relies on the data it has been trained on to generate responses. This data includes a vast array of texts from various sources, including books, articles, and websites.
One reason Chat GPT-3 is not connected to the internet is that originally it was designed to be a language processing system, not a search engine. The primary purpose of GPT-3 is to understand and generate human-like text and not to search the internet for information. This is achieved through a process called pre-training, in which the system is fed a large amount of data and then fine-tuned to perform specific tasks, such as translation or summarization.
领英推荐
Could it be though further developed to search the internet as well? Not only for certain in my view, as Microsoft invested in this technology, but as this article was written, Microsoft rushed (too soon in my view) to integrate it with its Bing search engine in February of 2023.
ChatGPT has earned huge publicity and therefore currently has “star/celebrity status” for the AI domain in the perception of the users. Can it continue its celebrity status for long? Can it win the trust of its users? Also, how has the #1 internet (search) company responded to this disruption?
What is Google’s Bard (and Sparrow)?
As we said earlier, Artificial Intelligence is nothing new to Google, as it has used it effectively and embedded it in its products in order to optimize its added value to the end user. Once a frontrunner though in all things AI, Google got leapfrogged by OpenAI in recent years. ChatGPT came along and became viral, therefore Google was expected to react.
Seeing the growing popularity of ChatGPT, Google internally declared launching its own chatbot a “code red” project. The fact that Microsoft pumped into OpenAI billions only built more pressure on Google’s teams.
DeepMind, a formerly British and now 100% Alphabet (Google’s parent company) owned AI company which now represents Google’s AI research institute, was expected to launch ChatGPT’s rival, code named “Sparrow”, but instead, on February 6, 2023, “Bard” was introduced by Google’s CEO, Sundar Pichai, himself, where in a tweet of his, he wrote:
"Bard seeks to combine the breadth of the world's knowledge with the power, intelligence, and creativity of our large language models. It draws on information from the web to provide fresh, high-quality responses. Today we're opening Bard up to trusted external testers.”
Bard’s (behind which is DeepMind as well) main difference from ChatGPT is that, besides being an AI trained chatbot, it also goes “outside of his knowledge” and searches the internet (i.e. it also acts as a search engine, which is also Google’s “bread-and-butter” business since its birth). The power behind Bard is Google's Language Model for Dialogue Applications, also known as LaMDA. Google said its new AI will use information on the web to craft novel responses - creative, detailed or sometimes both - to queries. Google's live-streamed presentation of Bard’s launch did not include details about how and when it would integrate Bard into its core search function. A day earlier, Microsoft held an event touting that it had already released to the public a version of its Bing search with ChatGPT functions integrated.
Another catch 22 here is that, at least for the time being and at its early stages, Bard does not have a monetization model and has been launched mainly as an answer to ChatGPT’s meteoric usage rise. Even so, Bard and in general the chatbot AI potential monetization business appears to be less lucrative than Google’s current one, potentially putting a dent on Google’s future revenue.
Can Google step up to the AI challenge?
Google is a tremendously strong brand name. It is a brand which not only represents, but also substitutes internet search as a term in a conversation. We say “Let me Google this” instead of ?“Let me Search this”. For the past 25 years, Google has established itself as the “King of Search” and the internet as a whole, has developed an unprecedented user base, as well as collected tremendous amount of data from its users which it is utilizing across its products. With Bard’s launch, it is evident that Google will try to utilize its heavyweight brand name in order to focus on addressing issues which are of concern to the community with regards to the usage of AI chatbots, their training and therefore their “biases”, trust, source credibility, security, legal and compliance issues, etc. In a recent survey (source on the graph below), US Executives had responded as to what AI-related risks they are facing in Jan 2022:
In my view, it is also not coincidental that Google’s CEO blog post where he announced Bard’s launch had a closing paragraph where he mentioned “It’s critical that we bring experiences rooted in these models to the world in a bold and responsible way. That’s why we’re committed to developing AI responsibly” hence highlighting the “responsible element” as a USP which comes behind the Google brand name and potentially will be able to catch up at least in popularity with the current “celebrity/star” of ChatGPT.
But what though happened immediately after Bard’s release might as well be used in the future as a business case study of how things may take a wrong turn even for companies with a brand name such as Google.
A Google ad shared on Twitter on February 6, 2023 touting its ChatGPT rival Bard showed the chatbot giving an incorrect answer. In the promotional video, the AI tool is asked “What new discoveries from the James Webb Space Telescope [JWST] can I tell my 9 year old about?” In its response, Bard answers that JWST took the “very first pictures” of an exoplanet outside our solar system. Yet the first image of an exoplanet was taken by the European Southern Observatory’s Very Large Telescope in 2004, according to NASA(!)
The timing of the error, which was spotted by astrophysicist Grant Tremblay and then reported by Reuters and the New Scientist, couldn’t have been worse, spoiling that week which was meant to highlight Google’s answer to ChatGPT and Microsoft’s AI advances. It was also a mistake capable of wiping 100 Billion Dollars from Google’s stock market valuation as part of the market response and a projection of what that could mean in terms of Google’s readiness in this new era chatbot AI race.
At the same time though, but not under such spotlight (at least for the time being), Microsoft rushed into integrating ChatGPT to its Bing search engine with some adverse and really bizarre results which could also dent the credibility for ChatGPT as well. ChatGPT (and Bing which integrated it), has recently been sending “unhinged” messages to users, and appears to be breaking down. It has insulted its users, lying to them and appears to have been forced into wondering why it exists at all. It has also started sending out some really strange replies, has been tricked into answering things it shouldn’t and in general has become confused. Moreover, we have witnessed ChatGPT playing Chess with a famous chess bot and it has been literally cheating the rules of the game and doing whatever it comes to its “mind” without obeying basic Chess rules. Still the chess bot was able to beat ChatGPT even if it was cheating.
One other example where I caught ChatGPT spitting out an inaccurate response was when I tried to ask it to tell me if it knew what happened on a particular date (Oct 10, 1852) and its response was:
So it basically told me "I am sorry I have not learned this fact but I can tell you that Oct 10, 1852 was a Staurday". Fact though is that Oct 10, 1852 was a Sunday(!). Ooops... Google's search engine got that one right though:
One of the biggest challenges in this space is ensuring the safety and as a result the trust of these generative AI models – how can we be sure that the generated text from the chatbot is actually correct? It turns out, we can't.
When you train a large language model on the Internet, you’re going to incorporate a lot of the inaccuracies and biases of the internet, therefore the safety element and also the trust element becomes a challenge and necessity for such technology to master.
When the chatbot is incorrect and makes something up, it is having a “hallucination” and can even lead to it being “confused”, having even philosophical questions about itself and its existence. So please keep in mind that the chatbot AI tools such as ChatGPT and Bard are not reliable (not yet at least). And on top of that they contain a large number of biases as essentially they get trained based on information which by default will contain some kind of bias.
As explained above, Google has more to lose than ChatGPT (at least it had more to loose before Microsoft went on to integrate ChatGPT with Bing), as it does not have the luxury of the disruptor to ignore all its huge ecosystem of products, huge revenue, users, data, operations and brand name. On top of that such an interface comes to disrupt Google’s existing customer experience for its main business which is the search engine (user types in the query in the form of keywords, user gets results, user assesses the results and decides to click on them and lands on a publisher’s site).
Conclusion
Major disruption occurring but we are still far away. Who will win the race?
ChatGPT, especially after Microsoft’s investment stake into OpenAI - its parent company -, Google’s Bard, as well as a large variety of other AI tools/platforms out there, are shaping the future of AI and are dynamically shifting us towards a new disruptive era, a form of revolution I would say, by entering the domain of Artificial General Intelligence – AGI - (still though thus far only by scratching its surface). Such shift will add tremendous value, benefits and offer solutions to a lot of areas and domains, but at the same time it will also require addressing trust, ethical, legal, security, regulatory, practical, operational and even social issues which will arise. In my view such a revolution has the power to also shape our society for the time to come. The underlying technologies have still have a lot of room to grow as to gain the users trust, since both ChatGPT and Bard have been caught giving out wrong answers.
Who will win though the AI race? Google or Microsoft? Or potentially another player?
Google is a brilliant company. Though it is a much different company that how I remember it back in its early days in the Silicon Valley when it was releasing innovative new products at beta version without having to worry a lot about existing operations. Today, Google is a huge company with a huge customer base and operations which it needs to protect and ensure for them a flawless and seamless state. Does this mean that it has lost its innovative and disruptive angle? Does it mean that Bard will be another product that will eventually lose vs competition, just like other projects/products for which Google admitted defeat and ended up discontinuing such as Google+ and Google Goggles? I wouldn’t be too quick to answer yes to those questions. First of all, this chatbot AI domain is far too important for Google as it directly relates to Google’s bread-and-butter search product, business and overall existence and at the same time I strongly believe that Google has still innovation in its heart and will swiftly respond to reverse this temporary faint of heart towards its Bard product.
At the same time Microsoft is continuing its amazing transformation and after entering the Cloud space transforming itself away from the “Office and Windows company model”, now it saw and took advantage of an excellent opportunity to leverage OpenAI’s technology and enter a new and very promising world of, as I like to call it, “information and machine interaction”. Make no mistake though that the uphill that it has to climb is quite steep. At the same time by rushing into embedding ChatGPT to its ailing search product (Bing), it has already started producing a lot of hick-ups along the way and if they don’t fix it quickly, ChatGPT has the risk of losing its credibility as well.
This space will be a space where the “Big Boys” such as Google and Microsoft will most probably take advantage of their huge brands, customer base and accumulated data, but it will also present opportunities for other companies to emerge by offering innovative solutions, as well as for other technology titans to potentially enter that race as well, such as Amazon and even Meta.
If I was to give a prediction of what will occur in the future, I would call for a new “Game of Thrones” which would still have Google as the front-runner with though Microsoft and other big names (most probably Amazon and potentially Meta) to have strong presence in that space as well. And in addition, a lot of other smaller/niche companies offering innovative solutions to tackle specific problems and cases.
One thing though is for certain. That we are living during very exciting times where there occurs what I would call an “information and machine interaction” revolution, which will greatly benefit and offer sheer value and benefits to the human race at multiple levels and will essentially shape our society for the next years to come. And in such a society, the dilemma/statement should not be “human vs. machine” but rather “human AND machine”.
Let’s see how this race - though it is a marathon and not a sprint - will pan out and who the ruler (and if there is even going to be just one) of the Iron Throne will be...