Tracing the AI family tree: say hello to ChatGPT’s grandmother
Siemens is one of the early pioneers of AI; its Central Research department was already working on the topic in the 1970s, long before many of today’s AI players were even founded. Already in 1974, an "interactive, automatic, natural language question and answer system" was to be researched - ChatGPT’s grandmother? A historical review: the road towards today's Industrial AI!
Takeaways
Siemens – an early AI pioneer
Despite today’s buzz, the research and application of Artificial Intelligence (AI) is not new. Basic principles were already explored in mathematical theory over one hundred years ago. The concept of AI itself was first proposed by British mathematician and logician Alan Turing in his 1950 paper "Computing Machinery and Intelligence". In 1956, a historic summer workshop brought together top researchers working on a program designed to mimic human problem-solving skills: held at Dartmouth College in New Hampshire, this is widely considered the birthplace and birthdate of AI as a field of study.
Fast-forward to today, where Siemens employs over 1,400 AI experts around the globe and has filed 3700 AI patent applications.? Ranked as the 13th company in the world based on the number of AI patent applications, Siemens is often considered to hold the second place worldwide among industry peers - this international leadership position didn't happen by chance. Siemens is one of the early pioneers of AI; its Central Research department was already working on the topic in the 1970s, long before many of today’s AI players were even founded.??
The 70s – science in an ivory tower?
For some, the 70s were considered the "AI winter". Optimistic expectations raised in the previous decades were struggling to be realized, as AI research seemed destined to remain in the academic ivory tower. Certain researchers, however, held strong and pushed forward, driven by belief in the technology’s practical potential.?
A very early vision and application of AI came from Siemens and its partners in 1974. Specifically, the Institute for German Language (Institut für deutsche Sprache, IdS) in Mannheim was supplied with the Siemens data processing system 4004/151, which was used in the "Linguistische Datenverarbeitung II" – a project sponsored by the German Federal Ministry of Research and Technology. An "interactive, automatic, natural language question and answer system" was to be researched to enhance database handling, as well as to improve human-machine communication. Could we consider this an ancestor of ChatGPT, possibly its grandmother?
Of course, one can only answer this question with a wink and a smile. But, looking at the following clipping, the early progress of AI back then, complete with bulky magnetic tape rolls, stands in stark contrast to today’s sleek smartphones.?
Meanwhile, the AI research at Siemens continued to pick up speed. In Siemens Magazine No. 53 (1979) an essay by Hans-Reiner Schuchmann (Zentralbereich Technik) appears with the title “Auf dem Wege zu einem neuen Computerverst?ndnis” (Towards a New Computer Understanding) in which "intelligent computers" are the focus subject.?
The 80s – a decade of expert systems
In the 1980s, further technical papers and lectures by scientists from Central Research appear, not least Egbert Lehmann’s (Zentrale Aufgaben Informationstechnik) publishings on expert systems (1984) and linguistics and AI (1985). There is even a historical film from the late 1980s by Siemens Television Service, which already tackles concrete applications of AI:
领英推荐
Watch the video below?and see the early AI systems at work!
Further areas of application also become apparent in these years, as an article by Siemens researcher Georg Eibl from 1985 points out. AI was already considered a core subject of computer science at this time, and this article details applications for CAD/CAM as knowledge-based expert systems as well as the AI programming languages LISP and PROLOG.?
The 90s – neural networks and strong partnerships
In 1994, Volkmar Sterzing and Oliver Mihatsch at Siemens started fruitful collaborations with the Massachusetts Institute of Technology (MIT) on reinforcement learning for admission control in telecommunication systems. Siemens and MIT – still enjoying a strong, strategic relationship today !
In the late 1990s and early 2000s, Siemens Corporate Technology researchers were among the first to harness artificial neural networks for innovative solutions: for example, the optimization of energy-intensive steel plants (Martin Schlang) and of large gas turbines (Volkmar Sterzing).?
And in 2001, Volker Tresp (Department of Technology, ZT) was awarded the Inventors of the Year prize for his “learning computers”, which were able to deal with incomplete data sets and complex interrelationships by using neural networks. Speeding up the neural networks with electro-optical computers was also part of his pioneering research.
2000s and 2010s - unprecedented opportunities
In these years, the application fields for AI seemed unlimited.?
A 2003 issue of Siemens’ Pictures of the Future magazine featured intelligent systems and neural networks for forecasting product sales with up to 85 percent accuracy, as well as a new route planning software 1,200 times faster than previous methods. Siemens opted here for "recurrent" networks instead of the "feedforward" networks most commonly used for sales forecasting. In the former, data flows in only one direction, from the input layer to the output. In contrast, in recurrent networks, signals from one layer are returned to the upstream layer, resulting in more robustness and the ability to train the network with less data. For the time, this was a highly innovative approach.?
Elsewhere, tangible, real-world impact was being realized: in 2006, around 60 rolling mills worldwide were already equipped with neural networks from Siemens.
All of which reinforces the fact that today's widespread attention towards AI in general, and large language models in particular, didn't emerge out of nowhere. We can liken it to the development of a wave; the gradual accumulation of energy, progress and transformative impact – the result of multiple forces and inputs converging. Building over time, ripples start to reach the surface of public awareness (via milestone, breakthrough developments) before the AI wave itself reaches the shores of society to transform the landscape. Now, with the figurative weather conditions optimal, it’s worth remembering that many brilliant minds have been diligently mapping and monitoring the progress of this wave for years – including many at Siemens.?
So, what’s next? With this historical review, you can glimpse just some of the potential of AI, not least within industrial processes today. It’s taken time to arrive but now we have the opportunity to look ahead, to harness its power and ride the wave as a strategic force for good. With this in mind, Industrial AI is the way Siemens does Artificial Intelligence. This is reliable and robust, secure and trustworthy; designed to meet the rigorous requirements and standards of the most demanding professional environments. This can help safeguard our future: managing and optimizing the complex, critical systems of our world that cannot afford to fail.
Postscript: This article does not claim to be complete in terms of Siemens’ AI history. The more you dig, the more you find. Much more undoubtedly remains hidden away in the archives and in the memories of experienced colleagues. If you have more details, developments or anecdotes concerning Siemens’ AI research, please do get in touch.?
With kind support of Siemens Historical Institute and Siemens AI Lab .
Supervisor at aman grafika
8 个月https://www.youtube.com/watch?v=7DfFdBut3BU
Experienced Quality, Production and Project management professional with knowledges of Marketing, Sales, IT, 3D modelling and AI Networks
1 年It is interesting what about current or planned AI products from Siemens. I actually actively use and study AI and their's application. Let me say - currently many companies are try to apply this revolutionary approach. For instance you can check Adobe max conference where you can see that AI will be integrated in many Adobe products. I think this can be really advantageous to apply AI tech in different industries and it is one of keys to success in future. And from my POV Siemens also should apply it and make it one of the most important directions otherwise in several years small lab that some years ago created Chat GPT will cost much more than entire Siemens itself.
Great read!
Project ochestrator
1 年Question: Did Siemens ever stop working on the subject? What was their strategy for not promoting their brand as a pioneer in the subject? Do they have any long-term plan to implant the feature as part of their products? (e.g. in automation/control systems)