OpenAI has released GPT-4.5 as a bridge to the upcoming GPT-5. The company continues to push the boundaries of large language models — GPT-4.5 brings significant improvements compared to the original GPT-4.0, including a broader context window, training on visual data, and the most up-to-date knowledge of the world. It is also claimed to possess much better emotional intelligence, more efficient code generation, and the ability to produce complex responses. Meanwhile, Anthropic's Claude continues to impress — the version known as Claude 3.5 "Sonnet" has established itself as an essential AI tool for coding, demonstrating superiority in programming tasks. The latest iteration, Claude 3.7 Sonnet, introduces the "Extended Thinking" mode, which enables the model to self-reflect during response generation, reducing the production of inaccurate information. These next-generation models highlight how quickly generative AI is evolving — in just a few months, we've seen more capable, contextually "intelligent" systems that better understand user intent and deliver more relevant results.

Midjourney Version 6 has made a massive leap in photorealism and the quality of generated images. Beyond language models, the progress in diffusion models for image generation is also evident. Stability AI's open-source model, Stable Diffusion 3.5, was launched with three size variants tailored to different needs. The most potent version, Stable Diffusion 3.5 Large, with 8.1 billion parameters, generates high-quality images up to 1-megapixel resolution, setting new standards in accurately following text prompts. An accelerated Large Turbo variant was also introduced, which, using Knowledge Distillation, can generate an image in just four diffusion steps — significantly faster than previous models. These upgrades aim to restore Stable Diffusion's dominance against increasingly intense competition like OpenAI's DALL·E 3 and the popular Midjourney. At the end of 2024, Midjourney — the favorite AI tool for designers and artists — released its Version 6, which, in its alpha phase, already showed a significant leap in photorealism and solved the long-standing issue of generating readable text within images. Midjourney V6 is the third generation of the model trained from scratch. It was developed over nine months on powerful clusters, resulting in realistic images that further blur the line between photography and AI art. This rapid development, from GPT-4 Turbo to Midjourney V6, reflects the dynamic nature of the AI ecosystem — models are becoming more capable, versatile, and ready for widespread practical application.

Businesses across various sectors are rapidly implementing artificial intelligence to optimize operations and drive innovation. Below is an overview of how AI is applied in banking, telecommunications, healthcare, the public sector, and retail, with specific examples and results.

• Banking: The financial sector is among the leaders in adopting AI. Generative AI is used for service personalization and automation, with 75% of banking executives believing that advanced AI applications will provide a competitive edge. Additionally, 50% of banks have already integrated generative AI into their products and processes. Large banks are making significant investments in AI capabilities — from acquiring top-tier NVIDIA GPUs to accelerate models to developing AI systems for fraud detection and chatbots for customer support.
• Telecommunications: AI plays an increasingly important role in telecom operators' network management and customer experience. By using machine learning algorithms to analyze traffic and network load, companies optimize 5G network operations and predict failures before they occur. Such predictive maintenance reduces unplanned outages and costs, while AI-driven traffic management dynamically adjusts capacity based on demand. Studies also show direct business impact — AI can increase telecom sales conversion by about 15% and reduce capital costs by approximately 10%. In customer support, over 53% of telecom companies already use AI chatbots and virtual assistants to communicate with customers, speeding up query resolution and improving customer satisfaction. The result is a more agile network and a personalized service offering with lower operational risks.
• Healthcare: AI is revolutionizing diagnostics and patient treatment in the healthcare sector. A recent example from the pharmaceutical industry shows that scientists, with the help of AI, discovered a new antibiotic after 60 years. This demonstrates AI's potential to accelerate drug discovery. Additionally, AI systems for disease detection (such as those trained to recognize patterns in radiological images) help doctors diagnose cancer and other diseases more quickly and accurately, increasing the chances of successful treatment. In operating rooms, AI-driven robotics assist in complex surgeries, enhancing precision and safety. Healthcare institutions are also implementing AI chat assistants for patient triage and answering frequently asked questions, alleviating the burden on medical staff and speeding up the delivery of information. The result is better treatment outcomes and more efficient healthcare systems that can handle more patients.
• Public Sector: Government institutions and public services are increasingly using AI to improve services to citizens and internal efficiency. According to the White House's inventory, 1,757 cases of AI application were reported in 2024 by 37 federal agencies in the United States, more than double the number from the previous year. The most common applications are in "mission-enabling" internal processes, healthcare, and public citizen services. Specifically, governments use AI to automate administrative tasks, predictive analytics in infrastructure maintenance (e.g., predicting when roads or equipment need repairs), and for smart cities — real-time traffic management, energy consumption optimization, and faster analysis of satellite images in emergencies. In security services, AI helps process vast amounts of data (such as video surveillance or cybersecurity) for quicker responses. Although the public sector often adopts new technologies more slowly, there is a significant increase in interest in initiatives to ensure responsible AI usage (such as ethical committees for AI and guidelines for algorithm transparency).
• Retail (Commerce): The retail industry uses AI to better predict customer behavior and optimize supply chains. For example, Walmart uses AI to forecast demand. AI models help optimize stock levels and avoid surpluses or shortages by analyzing historical sales data and market trends. According to McKinsey, this approach can reduce inventory costs by 10–40%, and Walmart reported savings of billions of dollars due to more efficient logistics and reduced waste. In terms of shopping experience, AI delivers personalized recommendations — online retailers like Amazon use algorithms that analyze customer habits and suggest products ("Customers who bought this item also viewed..."), which increases sales and customer satisfaction.
• Additionally, the concept of cashless stores is becoming a reality: Amazon Go stores are equipped with cameras and AI-driven sensors that automatically detect when a customer picks up or returns a product, tracking the contents of their basket in real-time. There is no traditional register at checkout — the AI system automatically charges the customer through the app. This "just walk out" technology, based on computer vision and deep neural networks, completely transforms the shopping experience, making it faster and simpler. Retailers integrating AI report more efficient operations, more precise consumer targeting, and greater customer loyalty, gaining a competitive advantage in the market.

Hybrid AI systems are becoming one of the key trends in the industry. This approach involves rule-based (logic-driven) and generative AI working together, combining the strengths of both paradigms – logical reasoning and neural learning. For example, rules or domain knowledge can set frameworks and checks for the output generated by large language models, resulting in more accurate and consistent results. These hybrid systems find applications ranging from medical diagnostics (where medical protocols guide AI analysis of symptoms) to personalized marketing (where business rules segment users and generative AI adapts content). Research shows growing interest in this field – 42% of surveyed developers report they are currently building hybrid AI solutions, suggesting these systems will increasingly influence real-world business cases.

At the same time, we are witnessing the accelerated integration of AI into existing software platforms. More and more software developers are embedding AI functionalities directly into their products – from office suites offering intelligent text completion and data analysis to CRM systems with built-in recommendations for the best actions to e-commerce platforms that automatically optimize prices and inventories. According to Gartner, by the end of 2024, generative AI will be integrated into 40% of business software, a massive increase from <5% in 2020. This "AI-ification" of software means users will more frequently have intelligent assistants and automated analytics within the tools they already use, making their work easier. For example, modern collaboration platforms now include AI assistants that can summarize long threads or suggest to-do lists from meetings, while document management systems use AI for automatic tagging and finding relevant files. Hybrid cloud environments also contribute to this integration – companies can connect their own data and tools with AI models through APIs, creating customized solutions (e.g., an internal chatbot trained on the company's knowledge base). This trend lowers the barrier to adopting AI – organizations don't need to build models from scratch but can leverage already embedded capabilities, speeding up digital transformation.

Another significant focus is multimodal AI models – models that simultaneously handle different types of data (text, image, sound, video). Traditionally, AI models have been specialized for one domain (e.g., NLP for text, CNN for images), but recent advancements combine these modalities into unified systems. GPT-4 was partially multimodal (text and image), and now the competition is pushing further: for instance, Google's Gemini 2.0 is designed to support multimodality. In February 2025, Google announced that Gemini 2.0 would be widely available via its Cloud, bringing developers the Multimodal Live API, enabling real-time two-way voice and video interaction with AI. This practically means that applications can use one model to interact with users via speech, understand their visual gestures or images, and generate responses in multiple formats (text, speech, and even image creation). Gemini 2.0 also introduces built-in image generation and advanced speech synthesis, so, for example, it can simultaneously edit an uploaded image and describe the changes through speech. Additionally, the so-called agent capabilities of the model have been enhanced, such as a better understanding of complex instructions, programming tasks, and calling external tools. All of this forms the foundation for virtual agents that can proactively execute tasks. In addition to Google, other companies are working on multimodal AIs: OpenAI will aim for AGI agents in future versions of GPT that combine multiple skills, Meta is exploring models that connect video, text, and audio, and startups are experimenting with AI that, for example, receives design requests via speech and instantly draws an app prototype. The multimodal approach is exciting because it brings AI closer to how humans perceive the world – integrated and contextual – which promises more intuitive interaction with technology and new types of applications (e.g., intelligent home assistants that understand and observe you or AI systems for monitoring production that simultaneously tracks visual and numerical data).

Capital investments in AI continue to grow almost exponentially, fueled by enthusiasm for generative models. Preliminary data shows that 2024 global VC funding for AI startups reached $131.5 billion, 52% more than the previous year. In comparison, funding for other startups fell by about 10% over the same period, meaning that artificial intelligence attracted a disproportionately large share of investments despite investor caution in other sectors. Even half of all venture capital investments in Q4 2024 were directed toward AI companies (50.8% share, compared to about 25% the previous year) – indicating that investors view AI as the next technological revolution they don't want to miss. Behind these numbers are mega investments in leading AI players: companies developing foundational models have attracted the most significant funding rounds. OpenAI secured $6.6 billion in October 2024, with an estimated valuation of $157 billion, solidifying its position as one of the most valuable AI companies in the world. Anthropic, the startup behind the Claude model, has also raised billions through a series of rounds led by Amazon, Google, and other interested giants.

At the same time, strategic partnerships and alliances are forming across the AI ecosystem. Companies recognize that by pooling resources and expertise, they can advance faster in developing AI solutions. A notable example comes from the telecom industry: South Korean giant SK Telecom invested $100 million in the American startup Anthropic to jointly develop a large language model tailored to telecommunications services. This collaboration expanded into a global alliance of several telecom operators (from Europe, Japan, and Canada), aiming to create a multilingual AI assistant for customer support and network operations that understands specific terminology and requirements for the industry. Big tech companies are also forming alliances: Amazon invested up to $4 billion in Anthropic at the end of 2023, ensuring that Amazon Web Services would be the primary Cloud for training their models. With this move, Amazon ensures that Claude models will be optimized for AWS infrastructure, attracting users to its Cloud, while Anthropic gains capital and resources to compete with OpenAI. Microsoft follows a similar strategy with its partnership with OpenAI – deep integration of the GPT-4 model into Microsoft products (Office Copilot, Bing AI search) and an exclusive Azure cloud contract. NVIDIA, a leader in AI hardware, is investing in software companies and startups to stimulate demand for its GPUs – for example, partnerships with medium-sized AI firms to optimize models on NVIDIA's platform. A consolidation trend is also visible: more prominent companies are acquiring AI startups to gain talent and technology (the so-called acquisition strategy). In retail, Walmart has acquired several AI firms for analytics to enhance its data platforms, while pharmaceutical companies have invested in AI startups for drug discovery. These networking and mergers highlight that the "AI race" is not only technological but also business-related – the faster one forms the right ecosystem of partners, the more advantageous it is for market product placement.

In the regulatory field, 2025 will bring significant changes that will shape the global AI market. The European Union is finalizing the adoption of the Artificial Intelligence Act (AI Act), the first comprehensive legal framework for AI systems. Although the complete requirements will take effect in 2026, specific provisions will start to apply on February 2, 2025 – companies developing or using AI in the EU must now take steps toward internal AI education (mandatory "AI literacy" for staff) and ensure that banned practices (such as social scoring systems or mass biometric surveillance) are not used. Failure to comply with these rules carries significant penalties, motivating firms to integrate mechanisms for ethical AI usage from the start. The United States has taken a different approach, combining government guidelines and voluntary commitments from the private sector. At the end of 2023, the US president issued an executive order requiring a range of measures: from establishing a Chief AI Officer function at the federal government level, monitoring the development of the most advanced AI models, to setting safety and privacy standards (e.g., ensuring that models do not produce biased or harmful content). At the same time, leading AI companies in the US (OpenAI, Google, Meta, and others) made a public commitment to the White House to extensively test their models for safety, label AI-generated content, and share information with the government for security checks. China is also actively shaping its regulatory landscape for AI: as early as mid-2023, it passed regulations requiring that every central generative model be registered with authorities and meet certain conditions (e.g., promoting "socialist values"), and since then, it has introduced rules for deepfake content and the use of personal data in AI. In practice, Chinese companies like Baidu and Alibaba now launch models and AI services only after obtaining regulatory approval, and the government is directly encouraging the development of domestic open models (such as the open Ziya 13B or Baichuan series) to reduce reliance on foreign technologies. It is expected that in 2025, China will further tighten quality and accountability standards for AI systems available to citizens. Globally, initiatives have been launched to coordinate AI governance: discussions are taking place in Geneva on a code of conduct for large AI models, while at a recent summit in Paris, a forum was established for exchanging best practices between governments and companies (focused on AI safety and ethics). All these regulatory activities bring more clarity and guidelines to the market but also potentially increase the costs of development and compliance, favoring more prominent players who can meet them. For users and society as a whole, stricter oversight promises more reliable AI systems and a lower risk of misuse.

The geopolitics of AI development is becoming an increasingly important topic, as technological advantage in AI can translate into economic and military superiority among nations. The US and China are at the center of this competition. The United States seeks to maintain its leadership position by limiting China's access to the most advanced hardware and tools – they have expanded export controls to now require licenses for exporting high-end AI chips to third countries and are introducing quotas for delivering high-performance chips to prevent redirection to China. In addition, US authorities are considering restrictions on the export of top-tier AI models or their weights, aware that software can be just as strategically significant as hardware. On the other hand, China is investing massive resources to accelerate self-sufficiency in AI – domestic companies are focusing on developing more efficient and cheaper models that can work on less powerful hardware and on open platforms accessible globally.

EGZAKTA Group recognized the importance of AI in a timely manner and made it part of its own DNA. During 2024, the company integrated artificial intelligence across all business segments, expanded its global presence, and redefined its role in the consulting industry. In its 2025 strategy, EGZAKTA focuses on bridging the digital gap in the Adriatic region through five complementary pillars: advisory services, technology, AI, investments, and education. This visionary approach, which combines investment in the latest technologies with knowledge transfer, positions EGZAKTA as the region's leading partner for digital transformation, capable of offering both strategic expertise and technical execution to its clients.

As a consulting partner, EGZAKTA helps organizations strategically implement AI and achieve tangible business benefits while minimizing risks. Our holistic approach combines a deep understanding of the client's business processes with the latest advancements in artificial intelligence. In practice, this means that we work with the client to identify areas where AI can deliver the most outstanding value (whether it's automating routine tasks, enhancing customer experiences through personalization, or creating entirely new AI-driven products). We then develop an AI strategy and roadmap aligned with the organization's goals, from selecting the proper use case scenarios to choosing the appropriate tools and models to plan the necessary changes in infrastructure and staffing. The EGZAKTA team provides end-to-end support, which includes pilot projects (where AI solutions are tested in controlled environments using real data), iterative model improvement based on user feedback, and scaling successful solutions throughout the organization. Understanding that trust and accountability must be at the core of any AI application; we pay special attention to risk management and ethical aspects. This includes evaluating potential biases in data and models (and mitigating them through data cleansing or fairness algorithms), ensuring compliance with regulations (whether it's GDPR requirements for data privacy or the new AI Act rules in the EU), and defining explicit usage and oversight rules for AI systems within the organization. EGZAKTA also advises on establishing AI governance frameworks—from forming internal AI and ethics committees to creating procedures for regularly assessing model performance and responding to unforeseen results. EGZAKTA Consulting aims to be the bridge between the vision and execution of AI initiatives: we help organizations move from the idea of artificial intelligence to successful implementation that delivers measurable results while safeguarding data integrity and the client's reputation. Through this partnership approach, our clients accelerate innovation while minimizing risks, ensuring long-term sustainability and competitive advantage in the age of AI transformation.

Summing up the week behind us, it is clear that the AI field is advancing at an unprecedented pace—from the launch of improved models and new hybrid solutions to significant investments and the beginning of regulations that will shape the future landscape. In the coming weeks, we can expect this momentum to continue. Major players are likely preparing new moves: OpenAI has already hinted that GPT-5 is on the horizon with potentially more autonomous "agent" capabilities and a deeper understanding of the world. Meanwhile, rivals such as Google-DeepMind, Anthropic, and xAI are rapidly iterating on their models (Gemini, Claude, Grok) in the race for dominance. We will also be watching how companies adapt to the new rules—especially in Europe, where the implementation of the AI Act sets deadlines for compliance, but also in other parts of the world where new guidelines and standards are expected (e.g., ISO standards for AI or initiatives like the Global Partnership on AI). The geopolitical development of AI could also bring surprises, whether through collaboration or further fragmentation of markets—such as whether the first agreements on limiting high-risk AI systems will be made on a global scale or if the tech rivalry between superpowers will intensify through new investment blocs. Either way, the AI ecosystem will continue to evolve at a rapid pace, bringing both new opportunities and new challenges. EGZAKTA will continue to monitor all these developments closely. In future editions of Egzakta AI Insights, we will aim to provide timely analyses of the most important trends—from technological advancements and industry applications to regulatory changes. Whether you're a business leader planning an AI strategy, an enthusiast following the latest models, or a professional concerned about ethical implications, our goal is to offer you a clear and concise overview of what's happening in the world of artificial intelligence. Thank you for being with us this week—we will continue to look toward the future of AI together!

Contact

Email: office@egzakta.com
Website: www.egzakta.com
LinkedIn: Egzakta Group and Egzakta Advisory

Follow us on social media for daily updates and exclusive analyses:

LinkedIn: Egzakta Group and Egzakta Advisory
Instagram: Egzakta Group and Egzakta Advisory
Facebook: Egzakta Advisory