Year: 2024

  • Girlsday 2024

    Girlsday 2024

    On Girls’ Day, young women have the chance to gain insights into the professional world for a day. A particular focus is on industries in which the proportion of women is less than 40%, such as IT, crafts, science and technology.

    Why do we actually need Girls’ Day?

    The explanation is simple. Despite particularly good schooling, more than half of the young women in Germany choose from only ten occupations in the dual system. Among these ten occupations, there is not a single scientific or technical one. Girls’ Day therefore offers young women the chance to immerse themselves in professional fields that they would otherwise not consider. It helps participating companies to present themselves to particularly female career starters and to dispel prejudices about certain activities.

    At the launch of this year’s Girls’ Day under the motto “Artificial Intelligence (AI) and the Transformation of the World of Work”, Federal Chancellor Olaf Scholz discussed STEM career prospects with female students and experts. The Federal Chancellor stated: “Girls’ Day is important, ‘as long as it takes’ – as long as necessary. Our society is diverse and different – this should be reflected in working life. That is why it is so important to encourage girls to take up technical, scientific or mathematical professions. And that is why so many large, small and medium-sized companies are involved in Girls’ Day. They know: good teams are usually mixed teams.”

    We fully agree with this statement and are supporting the campaign again this year by offering two female students the opportunity to take a look behind the scenes of Kontor Business IT GmbH and our subsidiary, Splitbot GmbH. We were very pleased to have the two 14-year-olds Sarah and Nila with us today and to show them our everyday working life. We hope we were able to show them that IT is not boring.

  • Our Split Blog in April – P versus NP

    How do you explain a topic to outsiders that your own mind can barely grasp? Our editorial team faced this challenge this month. We are attempting it and hope we have succeeded.

    Many thanks to our developer Max, who broadened our horizons with the topic suggestion “The P versus NP Problem”

    Before we delve into this specific topic, we first want to shed light on the term “Millennium Problem.”

    The term Millennium Problems currently refers to seven unsolved problems in mathematics listed in 2000 by the Clay Mathematics Institute (CMI) in Cambridge. The institute has promised a prize of one million US dollars for the solution of each of these problems, provided that they are published in a specialist journal.

    The list includes these problems:

    1. the proof of the Birch and Swinnerton-Dyer conjecture from number theory
    • Analysis of the existence and regularity of solutions to the initial value problem of the three-dimensional incompressible Navier-Stokes equations

    Only one of these problems, namely the proof of the Poincaré conjecture in topology, has been solved so far. The Russian mathematician Grigori Yakovlevich Perelman was able to prove in 2002 that the conjecture is correct. After three teams successfully verified the solution, Perelman was to be awarded the promised prize money in 2010, even though he had only published the solution on the Internet. However, Perelman declined the money and the associated award.

    Let’s now take a closer look at the P versus NP problem in computer science, which has been unsolved for decades. This is a so-called decision problem. The question that arises is whether the class of problems that can be solved algorithmically with relatively little effort (“polynomial time”) (the class P) is equivalent to the class of problems that, although not necessarily with little effort, can at least be checked with a reasonable effort (“non-deterministic polynomial”) (the class NP).

    To illustrate the problem, one can, for example, visualise the so-called “Knapsack problems” in computer science. Imagine you are planning a hike and need to pack your backpack. The question is: Which items fit in the backpack and how much weight can you still pack? You can easily answer this question by simply packing the backpack and checking the total amount of weight. However, the real challenge is to predict which items you should choose to achieve the optimal weight. This is a typical NP problem.

    The question posed by the Millennium Problem P versus NP is therefore: Can a computer solve an NP problem quickly and efficiently, or are these problems fundamentally more time-consuming, even for a powerful computer? Answering this question could revolutionise the world of computer technology.

    If it turns out that P equals NP, this would be a breakthrough in solving many difficult problems, including optimising machine learning and cryptography. It would even mean that the secrets of science and technology could be unlocked like never before and our world could be improved by the computing power of computers.

    On the other hand, answering the question that P is not equal to NP would mean that there is a fundamental limit to the computer world. Some problems are simply too complex for a computer to solve quickly and efficiently.

    That is why the Millennium Problem P versus NP is so significant and interesting. It is a challenge that tests the limits of mathematics and computer science and helps us to expand our knowledge and improve our world through the incredible power of computers.

    As we all know, a wide variety of AI systems are currently conquering the market. With regard to the problem described, in particular, AI could be the key. Perhaps an AI software will succeed in proving the assumption that P equals NP. Or one of the upcoming AI models itself could be proof that an NP problem can be solved just as quickly and efficiently by a powerful computer as a P problem.

  • AI.STARTUP.HUB Hamburg – We are in!

    AI.STARTUP.HUB Hamburg – We are in!

    In a startup, there is always something new to learn. New goals are to be achieved, new challenges to be mastered. But what is the best way to approach this? Where can one find reliable expert knowledge, support, assistance, and indispensable contacts? Our recommendation: participate in an accelerator program. After gaining invaluable experience at the Lübeck accelerator program Gateway49, we have now successfully joined the Summer Batch 2024 of AI.STARTUP.HUB Hamburg.

    The AI.STARTUP.HUB offers excellent support to (aspiring) AI founders and their teams throughout all development phases. From ideation to internationalization – every team here has the opportunity to learn from the best in the industry.

    To cater to as many teams and development stages as possible, the AI.STARTUP.HUB offers two different programs: The AI Accelerator Program for already established startups and the AI Ideation Program for teams that aim to launch successfully and bring an innovative idea.

    The AI.STARTUP HUB Hamburg is a safe haven for all who are ready to actively contribute and develop further. The Hub is a joint project of Hamburg Innovation GmbH and the Artificial Intelligence Center Hamburg (ARIC) e. V. and is implemented together with partners Exponential Innovation Institute (Exii), Machine Learning in Engineering (MLE) of Hamburg University of Technology (TUHH), and German Entrepreneurship. Such a concentration of expertise in one format is a boon for any startup.

    Besides coaching and mentoring, one factor was particularly crucial for our participation: the exchange with other teams from the industry. The value of broadening one’s perspective should not be underestimated. Therefore, we are especially looking forward to sharing both our successes and our failures with the other teams, thus learning from and with each other.

    Our six-month journey has now begun with the kickoff on April 9 at the new location at Sandtorkai in Hamburg. It was a fitting start with a total of 17 exciting pitches in an incredibly relaxed and constructive atmosphere. We are enthusiastic about the other teams and the mentors and very proud of this further milestone on our journey.

  • Our Splitblog in March – Deep Mind Gemini 1.5

    Today we want to focus on the new AI model from Google. This is a multimodal AI model that can process various types of information, such as texts, images, program codes, and audio information and their combinations.

    A topic suggestion from our developer Mats, who is primarily responsible for the development of our chatbot Kosmo

    A few weeks ago, Google introduced DeepMind Gemini 1.5 – an update to the previous AI models from Google.

    The amount of data that Gemini 1.5 can process is particularly groundbreaking. Up to 1 million tokens can be provided in the context window. In internal experiments, the amount of data could even be increased to 10 million tokens. A token is a kind of basic unit with which, for example, sentences can be divided into smaller units (tokens) and thus processed by the model. A token is therefore a group of characters. For comparison: Chat GPT-4 Turbo can process 128,000 tokens (as of December 2023). This is roughly equivalent to a 300-page book. If more pages were provided, the model would no longer be able to access the information on the first pages. Figuratively speaking, at the end of a book, it would no longer know the author’s name.

    Gemini 1.5 can capture and analyze up to one hour of video material, eleven hours of audio recordings, texts with up to 700,000 words, or 30,000 lines of code. And, what is even more amazing, it can “remember” the content and connect it with new information.

    During the presentation of the new model, Gemini 1.5 was tasked with analyzing the 402-page transcript of the Apollo 11 mission and finding three humorous passages in it. In fact, the model succeeded in identifying three entertaining moments within about 30 seconds. For example, Command Module Pilot Michael Colins said at one point: “The Tsar is brushing his teeth, so I’m stepping in for him.”

    Without further information, the researchers then uploaded a hand-crafted drawing of a leaking boot and asked which moment was shown in the picture. The answer came promptly: “One small step for a man, but one giant leap for mankind.” Gemini 1.5 can therefore establish complex relationships and reproduce them correctly without concrete instructions.

    1. The architecture of the model is also advanced. It is no longer a uniform, large model approach, but a collection of smaller, specialized transformer models. This type of architecture is called Mixture of Experts (MoE). Each of these transformer models is, so to speak, an expert in its field and able to handle certain data segments or different tasks. Based on the incoming data, the most suitable model for the application is dynamically selected. For different inputs, different sub-networks of the model are activated for the appropriate outputs.

    This approach enormously increases the efficiency and quality of the results delivered.

    Gemini 1.5 is currently only available to selected corporate customers and developers. We are excited about the further development.

  • Is ChatGPT without Alternatives?

    Is ChatGPT without Alternatives?

    As part of the series ‘Success is the best job security,’ the ‘AI for SMEs’ event took place on Tuesday at Media Docks Lübeck. Dirk Schrödter opened the event with an appeal to the approximately 175 entrepreneurs present, advocating for courage in adopting new technologies. To remain successful and competitive in the long term, reservations must be overcome.

    Bodo Neumann from Grafix also encouraged those present to explore meaningful applications of AI within their companies.

    With the aim of familiarizing guests with AI applications, Caro and Tadeusz were on site, ready to answer any questions from interested parties. In a presentation titled ‘Overview of Open Source AI Solutions and How to Implement Them,’ our CEO took the opportunity to introduce the audience to solutions beyond the mainstream and discuss their GDPR-compliant application possibilities.

    Tadeusz compared various open-source and closed-source applications, detailing their respective advantages and disadvantages. It is often mistakenly assumed that open source is insecure and not GDPR-compliant compared to other applications. However, the opposite is true: with open source, users secure their digital sovereignty and absolute data security. Open source software offers companies the option of operating it in their own data center and is also individually adaptable to specific needs. Since open source models can draw upon a worldwide community, their performance is usually many times higher.

    • Further advantages of Open Source:
    • Transparent codebase
    • Collaborative development
    • Free of license fees
    • Open interfaces
    • Traceable development process
    • More innovation through open exchange
    • Access to expert knowledge

    Afterward, many interesting discussions arose, allowing for further exploration of the topics. It was an extremely successful evening, and we thank the Business Development Agency Lübeck and the Merchants’ Guild of Lübeck, and especially the curious and innovation-minded audience, for the magnificent event.

  • Our Splitblog in February – AI as an Opponent in Video Games

    Our Splitblog in February – AI as an Opponent in Video Games

    How do you fill a blog from the world of AI with exciting and interesting content? So many topics have been discussed in detail by so many people in recent months. What new aspects can we add to it? We asked ourselves these questions in the team today and once again it turned out that the spontaneous ideas are often the best. A new blog section, the “Splitblog” is born. From now on, every month a team member can choose a topic from the AI environment that will be highlighted in this category.

    It starts with our backend developer Florian and the topic “AI as an opponent in video games”

    A brief look at the video games of recent years shows that most of the improvements have largely been related to the graphics. More and more details, ever larger worlds, ever better resolution. The graphics in many games are now so sophisticated that they can hardly be distinguished from reality. But what about the behavior of the various characters? Especially when it comes to programming the NPCs, it is reasonable to assume that artificial intelligence has been used here for a long time. NPCs are the non-playable characters in a video game, such as passers-by, road users or even opponents. In many games, it is already possible to interact and communicate with them and their behavior often seems unpredictable. But what looks like artificial intelligence from the outside is actually the achievement of the game developers. Instead of artificial intelligence, video games often use so-called “pathfinding”. This means that the paths and actions of the NPCs were determined during the development of the game. The options of the NPCs are limited in this case and can depend on various factors, such as the strength of the player. The more possible options and factors that are defined in the development, the more realistic the behavior of the NPCs appears.

    But why isn’t AI used in game development?

    The assumption is that the use of artificial intelligence improves the gaming experience. The behavior of the NPCs would be more situational, individual and less predictable. Communication could also be adapted much more to the players.

    What is often disregarded in this supposed improvement, however, is that a learning AI could quickly ruin any fun. The probability is high that it would be almost impossible for the players to win against AI-controlled NPCs. Also, because the NPCs could team up.

    And another factor would be difficult to control: NPCs that are based on AI would, for example, be able to leave locations and simply “no longer participate”. Under these conditions, storytelling within a video game is not feasible.

    In simple terms, the use of artificial intelligence in video games would negatively affect the fun of the game. Who likes to play a game that you can’t win?

    Nevertheless, there are first games in which artificial intelligence is used. The worlds and possibilities are still limited in these games, but that will change. Development teams have already succeeded in limiting the superiority of AI-controlled NPCs, for example. A good example of this is AlphaStar. An AI-based program that has been trained to play StarCraft II. Here it has already been possible to throttle the AI in its actions and abilities, so that AlphaStar remains a fair opponent (at least for absolute game professionals). We can be curious to see how AI can be integrated into video games in the coming years.

  • Year in Review 2023

    Year in Review 2023

    A momentous year is drawing to a close. We are almost at a loss for words considering all the incredible events of the past months, but we nevertheless wish to take this opportunity to reflect on 2023.

    Those who have followed Splitbot’s journey know that Kontor Business IT GmbH has been working on AI-based software since 2021. This software automates (administrative) processes in various scenarios and returns a crucial work factor to skilled professionals: time.

    The urgent needs of our customers and the consistently open-minded and positive feedback then tipped the scales for a significant corporate decision. On March 15th of this year, Splitbot GmbH was spun off as a wholly-owned subsidiary of Kontor Business IT. This milestone not only highlights Splitbot’s potential but also enabled immense growth. From originally one developer working on the project with Tadeusz, a team of ten people has emerged, working exclusively for Splitbot. Additional colleagues undertake important tasks for both companies, thus ensuring success. Core team members and the minds behind Splitbot are: Florian Roßmann, Mats Kastner, Søren Molkentin, Bartosz Golis, Maximilian Esch, Muhemd Al-Moayad, and Katharina Kirstein. We are particularly proud to be able to support two young individuals in their training: Arturs Tinte and Ramtin Abouie. Vincent Schiller has also recently joined the team as a working student.

    Anyone who has visited our website is also familiar with the faces of our bots. Behind them lies the creative and design work of Friedrich Wehrmann and Maximilian Hertwig. We are also particularly grateful to Kristina Andresen, who, with tireless dedication, handles all administrative tasks and thus made a significant contribution to the creation of Splitbot. Last but not least, Carolina Wehrmann and Tadeusz Nikitin should, of course, be mentioned here. Without them as management, and without their belief in the project and their courage to break new ground, none of this would have been possible.

    However, mere presence is not enough for a company’s success. Every individual contributed to our ability to achieve further significant goals.

    A significant phase for us was particularly our participation in the Gateway 49 Accelerator Program. During this time, we learned an incredible amount and were able to build a broad network. We are very grateful for the support of the entire Gateway team.

    We were greatly honored by the visit of State Secretary Julia Carstens. Ms. Carstens was given a detailed presentation by us on Splitbot’s capabilities and interacted with Sigma in our VR room.

    A great success for us was also our inclusion in the Alpha Program of this year’s Web Summit in Lisbon. We were not only able to participate in this enormous event but also to present Splitbot at an exhibition stand, in the Startup Showcase, and in a 40 Words video. We are still overwhelmed by Lisbon and all the impressions we were able to gather.

    Through collaboration with DiWiSH, the AI Bundesverband, KI.SH, and WTSH, we were not only able to participate in various events but also to introduce ourselves and actively shape the digital future. Thus, Tadeusz, with his AI expertise, has become a sought-after speaker and a valued contact person at all levels.

    The successful deployment of Splitbot in our administration and in Business IT support were further events we were able to celebrate. The onboarding of our pilot customers is progressing successfully, and we will begin the new year with the implementation of further customer projects. New and promising ideas are already awaiting implementation, and we are excited to see where our journey will take us in 2024.

    Until then, we wish you and your loved ones wonderful holidays and a peaceful New Year.

    The Splitbot GmbH Team