Is AI Overhyped?

AI is becoming increasingly embedded in many of the things we interact with and use on a daily basis. AI is becoming increasingly more advanced over time, with some remarkable capabilities emerging over the past few years. There are many real-world, practical examples organizations and governments are using AI for their day-to-day activities. So, AI is a real thing, right?


Well, on the other hand there’s still substantial hype when it comes to AI. It’s easy to get wrapped in that hype when we hear about companies and VC firms making staggeringly large investments into AI research.

Are these big-dollar investments going to result in expected returns for the investors and actually push the field forward? Is the hype adequate given AI’s successes? Or are we in just the latest bubble and wave of AI interest? Is AI simply too overhyped?


What’s actually going on with AI adoption and project

The concept of AI goes back decades, with development coming and going in waves. In this latest wave, AI is enjoying mainstream success with companies of all sorts investing in AI, from the smallest of firms to the largest companies and governments. However, unlike the way that science fiction movies and books would like to portray AI, most of the AI implementation today is mundane or “boring” in nature. Rather than seeing lots of autonomous machines roaming the planet, the most successful AI implementations are more snoozy affairs such as document classification or advanced predictive analytics solutions, but they are providing real value to those using it. 


While these are real-world applications solving day-to-day problems, the problem is that “boring” doesn’t sell news. So many of these boring but highly valuable use cases and applications are not being discussed and written about. Instead, headlines talking about AGI, and robots that are humanoid, and other science fiction style applications are the ones people read. This further enhances the general view that AI is all about the autonomous pattern of AI. This pattern however is one of the hardest to implement as it’s completely removing the human from the loop and takes a long time to see ROI. 


On top of this, the continual bold promises made by some AI companies often remain undelivered. Where are our autonomous vehicles? Where is the super intelligence that’s supposedly around the corner? Why are the voice assistants still so dumb? While narrow applications of AI are moving ahead with gusto, these more significant challenges for intelligence still stand in the way. Rather than seeing rapid proliferation of autonomous systems, the most widely implemented solutions are those of the augmented intelligence  nature in which the human remains in the loop. They might not be as interesting as the science-fiction style implementations grabbing headlines and creating hype in AI projects. 


Is the vendor environment cooling off? 

For the past few years, AI has been a hot area for investment. For perspective, in 2010 the average early-stage round for AI or machine learning startups was about $4.8 million. However, in 2017, total funding increased to $11.7 million for first round early stage funding, a more than 200% increase, and in 2018 AI investment hit an all time high with over $9.3 Billion raised by AI companies


With all this money being pumped into startups, it’s only natural that VCs want to see a return on that investment. As a result, we’re starting to see some consolidation in the whole AI and automation marketplace from large and small companies alike. In May 2019 HPE acquired supercomputing leader Cray. In June 2019 Mighty AI, a company focused on creating training data for computer vision models for autonomous vehicles was acquired by Uber’s Advanced Technologies Group to help the company with their push into self-driving cars. In May 2020 Microsoft acquired RPA vendor Softomotive to help expand low-code robotic process automation capabilities in Microsoft Power Automate. DataRobot, a startup in the AutoML space, recently acquired both data prep firm Paxata and MLOps startup ParallelM. These are just a few of the many acquisitions that have been taking place.


We’re also seeing some interesting things in the area of robotics. Some robotics companies such as Boston Dynamics or iRobot seem to be thriving. While other companies in their industry are struggling to get proper investments despite producing respectable and reliable products. Firms including Rethink Robotics, Jibo, Anki, Mayfield, Reach Robotics, Starsky Robotics, Google’s Schaft, and many others have shut down completely. While software “robotics” vendors continue to raise billions of dollars as an industry, physical robotics firms struggle to make ends meet. Why is it that some robotic companies are doing just fine while others can’t get funding needed to survive?


Not too long ago, AI was so incredibly hyped that just the mention of the technology in any marketing material or pitch decks would get the company funding or customer interest. However, the tide might be changing. Companies are being forced to prove their AI capabilities and customers and investors alike are getting smarter in the questions they ask and making companies prove their AI capabilities rather than taking them at their word. Additionally we are seeing acquisition in the space to provide companies a wider range of capabilities rather than being hyper focused in niche areas.


Is Artificial General Intelligence (AGI) around the corner?

Part of the challenge of AI is that it still means different things to different people. In conversations about AI, the topic can be about a wide range of diverse patterns from autonomous systems to chatbots to predictive analytics. After all, AI enabled chatbots are just as much AI as self driving cars or a hyperpersonalized ad. However, these practical aspects of AI are not as interesting as the more sci-fi ones given to us thanks to Hollywood. When you ask a typical lay-person about AI, they think of the Terminator or Westworld. This concept is not the narrow AI of today’s practical implementations, but rather the idea of the Artificial General Intelligence (AGI) — a super-intelligent machine that acts and thinks and looks just like a human.


Companies that are striving to create a type of ‘super-intelligence’ are those who often receive the most investment despite there being no guarantee that anything will come of the research. It is then when the companies advertise their intents and plans without ever following through that leads AI to its overhyped reputation. We are far away from ever creating such a machine, yet that doesn’t stop the hype around this idea. 


In 2019, Microsoft made a huge investment of $1 billion in AGI research company OpenAI. The investment was made with the goal of advancing research in developing practical AGI systems. Since its investment, Microsoft says it has created one of the world’s top supercomputers for the exclusive use of OpenAI to work towards their goal of AGI. This supercomputer will give OpenAI the compute power and resources it needs to train powerful new AI models.


Is AGI really around the corner, or are we chasing an elusive goal that we may never realize? Is AGI even the desired end goal for AI, or are these more narrow applications better served to our needs? Will we find ourselves in a new AI winter when the promise of AGI is not met, and the narrow applications of AI are no longer even considered to be AI?


For many organizations, the chase for AGI is not that important. These companies implementing AI here and now see it as more important to make sure to focus on the practical and real-life aspects of AI. There are companies and governments out there right now doing some incredible things with narrow and not-as-hyped corners of AI. From automatic document processing, to personalized loans, intelligent automation, or AI enabled chatbots or voice assistants, these real-world applications already exist and are providing real value. Focusing on these applications, rather than the far-off dream of AGI, is the best option for those interested in the industry. Perhaps one day the hype around AI will go away. But it is hoped that the practical applications are here to stay.


Source: paper.li

Recent Posts

See All

Build simple fuzzer - part 1

First of all, we are learning here and this fuzzer is in no way going to be a proper tool used against real targets (at least not initially). This is why we are going to code it in python. For real to

Build simple fuzzer - part 2

In the previous part of this mini-series we’ve implemented a very simple fuzzer. As the main idea behind it is being an exercise therefore I don’t think it is capable of finding bugs in complex target