What is Quantum Computing good for?

When it comes to quantum computing (QC), after the quite real breakthroughs in hardware and some spectacular announcements under titles like “Quantum Supremacy“, the usual hype cycle has developed with a phase of vague and exaggerated expectations. I would like to briefly outline here why the enormous effort is being made in this area and what realistic expectations lie behind it.

To understand the fundamental differences between QC and Classical Computing (CC), we first need to take a step back and ask on what basis both computing paradigms operate. For the CC, the basis is the universal Turing machine expressed in the ubiquitous von Neumann architecture. This may sound a bit outlandish, but in principle it is easy to understand: An universal Turing machine abstracts the fact of programming any algorithm into a classical computer (universal) that is somehow (classically) algorithmically expressible (Turing machine).

The vast majority of “algorithms” that are implemented in practice are simple sequences of actions that react to external events such as mouse clicks on a web page, transactions in a web store or messages from other computers in the network. A very very small, but important, number of programs do what is generally associated with the word algorithm, which is to perform arithmetic operations to solve a mathematical problem. The Turing machine is the adapted thought model for programming these problems and leads to programming languages having the constructs we are used to: loops, branches, elementary arithmetic operations etc.

What is the computing paradigm for a quantum computer?

A quantum computer is built up of quantum states that can be entangled with each other and evolved via quantum gates. This is also a bit off the wall, but simply means that a quantum computer is set to have an initial (quantum) state that evolves in time and is measured at the end. The paradigm for a quantum computer is therefore the Schrödinger equation, the fundamental equation of quantum mechanics. Even without understanding the details, it should be clear that everyday problems are difficult to squeeze into the formalism of quantum mechanics and this effort probably does not bring any profit: Quantum mechanics is just not the adjusted model of thought for the most (“everyday”) problems and it is also not more efficient in solving them.

So what can you do with it?

The answer is very simple: QC is essentially a method for quantum computing. Now this sounds redundant, but it means that a quantum computer is a universal machine to calculate quantum systems. This vision, formulated by Richard Feynman way back in 1981, is still followed by the logic of research today. Thus, it is not surprising that publications on the subject dealing with applications are located either in quantum chemistry or in the basic research of physics [5][6].

Why does this matter?

Because the classical computer is very inefficient in calculating or simulating quantum systems. This inefficiency is basically due to the mathematical structure of quantum mechanics and will not be solved by classical algorithms, no matter how good they are. In addition to basic research issues, QC is likely to become important in the hardware of classical computers, where miniaturization is pushing the limits of designing transistors on chips using classical theories of electricity. 

Besides, there are a lot of interesting connections to number theory and other various problems, which so far can be classified as interesting curiosities. Based on current knowledge, the connection to number theory alone could have a significant impact, because for historical reasons almost all practical asymmetric encryption schemes rely on algorithms that essentially assume (there is no proof) that prime number factorization cannot be solved efficiently with classical algorithms. Quantum computers can do this in principle but are far away from being able to do so in terms of hardware.

AI – Where we are in the Hype Cycle and how it continues

While the artificial intelligence index shows that the increase of research articles and conferences in the field of AI continues, the media is slowly showing some fatigue in the face of the hype. So it’s time to take stock: What has been achieved? What is practically possible? And what is the way forward?

What has been achieved?

In the years 2018 and 2019 the previously developed methods for the application of neural networks (this is how I define AI here) were further refined and perfected. Whereas the focus was initially (2012-2016, Imagenet competition) on methods for image classification and processing and then on audio methods (2015-2017, launch of Alexa and other language assistants), major advances in text processing and generation were made in 2019 (NLP = natural language processing). Overall, the available technologies have been further improved and combined with a great deal of effort, especially from the major players (Google, Facebook, OpenAI, Microsoft).

What is practically possible?

The use of AI is still essentially limited to four areas of application:

  • Images: image recognition and segmentation
  • Audio: Conversion from speech to text and vice versa
  • NLP: word processing and generation
  • Labeled Data: Prediction of the label (e.g. price) from a set of features

This list is surprisingly short, measured by the attention AI receives in the media. The most impressive successes of AI, however, result from a combination of techniques such as speech assistants using a combination of audio, NLP and labeled data to convert the input into text, recognition of text intention with NLP and prediction of the speaker’s wish by using huge amounts of labeled data, meaning previous evaluations of similar utterances.

Decisive for the development of precisely these AI application fields were

  1. the existence of large quantities of freely available benchmark data sets (data sets for machine learning) on which algorithms have been developed and compared
  2. a large community of researchers who have jointly agreed on the benchmark data sets and compared their algorithms in public competitions (GLUE, Benchmarks AI, Machine Translation, etc.)
  3. a free availability of the developed models, which serve as a starting point for the practical application (exemplary Tensorflow Hub)

Based on these prerequisites one can quickly assess how realistic some marketing fantasies are. For example, there are neither benchmark data sets nor a community of researchers for the often strikingly presented field of application of predictive maintenance, and accordingly there are no models.

What’s next?

On the one hand, it is foreseeable that the further development in the AI area will certainly continue initially in the above-mentioned fields of application and continue to develop in the peripheral areas. On the other hand, areas are emerging which, similar to the above-mentioned fields of application, will be driven forward at the expense of large public and private funds (e.g. OpenAI and Deepmind are being subsidised by Elon Musk and Google with billions of euros respectively). An example of large investments in this area is certainly autonomous driving, but also the area of IoT. In total, I see the following areas developing strongly in 2020-2022:

  • The combination of reinforcement learning with AI areas for faster learning of models
  • A further strengthening in the area of autonomous driving resulting from the application and combination of AI and reinforcement learning
  • Breakthroughs in the generalization of the knowledge gained from image processing to 3D (Geometric Deep Learning and Graph Networks)
  • A fusion of traditional methods from statistics with neural networks
  • IoT time series (see below)

I see a big change coming with the rise of IoT and the associated sensor technology and data. By their very nature, IoT data are time series that must be filtered, combined, smoothed and enriched for evaluation. Relatively little specific has been done to date for this purpose. It could be that from 2020 – 2022, this topic could hold some surprising twists and breakthroughs for us. German industry in particular, which has benefited rather little from the initial developments in the field of AI, should find a promising area of application here.

Are data science platforms a good idea?

According to Karl Valentin: Platforms are beautiful and take a lot of work off your neck. The idea of platforms for automatic data analysis comes at just the right time. In line with this, Gartner has now published a “Magic Quadrant for Data Science and Machine Learning Platforms“. The document itself can only be viewed behind a paywall, but on the net some of the companies mentioned in the report offer access to the document by entering the address.

Gartner particularly emphasizes that such a platform should provide everything you need from a single source, unlike various individual components that are not directly coordinated with each other.

Sounds good to me! However, data science is not an area where you can magically get ahead with a tool or even a platform. The development of solutions – for example, for predictive maintenance of the machines offered by a company – goes through various phases, with cleaning/wrangling and preprocessing accounting for most of the work. In this area, ETL (Extract, Transform, Load) and visualization tools such as Tableau can be ranked. And beyond the imaginary comfort zone of platforms that managers imagine, database queries and scripts for transformation and aggregation in Python or R are simply the means of choice. A look at data science online tutorials from top providers like Coursera underlines the importance of these – well – down-to-earth tools. “Statistical analysis, Python programming with NumPy, pandas, matplotlib, and Seaborn, Advanced statistical analysis, Tableau, machine learning with stats models and scikit-learn, deep learning with TensorFlow” is one of Udemy’s course programs.

In addition, the projects often get stuck in this preliminary stage or are cancelled. There are many reasons for this:

  • no analytical/statistical approach can be found
  • the original idea proves to be unfeasible
  • the data is not available in the quantity or quality you need
  • simple analyses and visualizations are enough and everything else would be “oversized”.

This is no big deal, as it only means that the automated use of Machine Learning and AI does not make a data treasure out of every data set. If, however, the productive benefit becomes apparent, it is necessary to prepare for the production pipeline and time or resource constraints. Usually you start from scratch and reproduce everything again, e.g. in Tensorflow for neural networks or in custom libraries.

The misunderstanding is that a) Data Science can be driven up to productive use without a trace and b) a one-stop-shop for Data Science (here “platform”) is needed that does everything in one go. That will never happen.

This is really good news, because it means that organizations can achieve their first goals without having to resort to large platforms. The reasonably careful selection of suitable tools (many of them open source) helps to achieve this.

Also interesting:
In my video “AI Needs Strategy” I explain which steps companies can take to to use AI technology in a successful way.