I'm Anna Litrakenenson. I'm chief AI product officer and founder at Altech AI. And I teach AI product development and management at Georgetown University at their AI management master's program.I I've been fortunate to see the development of AI in the last, two decades. I, did my masters in computer vision and PhD in applied mathematics at the Weizmann Institute of Science. And since then, I spent, nearly twenty years developing AI powered products across industries. These are some of the companies I worked at or for, helping them developing, AI products and AI strategy, AI prototypes, and leveraging AI through their business. So let's kind of set our terms straight as we begin here.

AI is really an umbrella term for machines mimicking the way humans learn, humans reason, and even, humans do how humans do actions. And most of what we talked talk about in AI is really machine learning. And just to kind of set terms straight, let's say that if we're talking about the application side of things, we will say AI. And if we're talking about the model side of things, we will say machine learning. But most of the AI that you see, use, and familiar with today is really powered by machine learning models. And there are many types of machine learning models, simple statistical models, heuristic models, black box models.

But most of the models that we're using today are really deep learning models that started the flourishing in its 20 tens. And deep learning models are artificial neural network models with many layers. Hence, they are called deep. And they try to mimic the way the human brain works. Generative AI, which is AI that generates text, audio, video, pictures, code, is really most of the generative AI that created the current hype. And, it the generative AI, which which generated most of the current AI hype in our world is really powered by deep learning models mostly. And large language models, which, the heart of generative AI, are part of the larger field of natural language processing that's been also flourishing in the last fifteen years. GPT, which is, generating a generative pre trained transformer, is, a type of large language model and part of the generative AI scheme.

And a fun fact is that transformers were actually invented by a group of Google researchers. So was AI invented in 2020 with all the hype that was generated by releasing ChatJupyTE in November 2022? As most of you probably know, not really. AI is about 70 years old. And and it had huge advancements over the years since it started in 1950. So, really, we can attribute the birth of AI to Alan Turing's paper, computing machinery and intelligence, when he posed the questions, can machines think? That led to the Turing test. Initially, it was called the imitation game, and the Turing test is really whether a human can decipher between a human and AI in a conversation.

Today, first time ever in history, we're close to having multiple systems passing the Turing test. The the term AI was kind of birthed in the at the Dartmouth Conference in 1956 that was organized by John McCarthy, and AI as a research field started kind of then. In the sixties, seventies, and eighties, we had a lot of experimentations, and and we we had a flourishing of expert systems that were heuristic systems that were called artificial intelligence or AI. And something I really wanna mention is Aron Coyne's Aron. Aron Coyne was an artist and a technologist who probably built the first ever generative AI system before the term was even existing. Earl Harold Cohen, who was a painter, trained, software and also physical machine that then generated painting in his style.

So he trained the system with his painting and then had the machine generate paintings in his style. He done it gradually. First, it was just line shapes, and then he added people and and and plants. He started the system in the seventies that early, and he continued developing it through the decades until he died in 2016. I highly recommend checking Harold Klein's and Aaron's work. Harold Klein talk about it as machine and human boosting each other's creativity by this collaboration and brainstorming. And it is really fascinating that even then, it was possible to enhance human creativity with AI. In the nineties, we had the flourishing of big data and machine learning. In February, AI started being in our everyday life. Think spell checkers. Think Amazon. Think recommendation engines. Think Netflix and other streaming, like, music streaming, for example.

But it all kind of really started to take off with the deep learning invention in the twenty tens because that started to take big leaps thanks to the advancements in computing and the use of GPUs. And many companies started leveraging deep learning AI for their internal operations, business automation, predictive systems, analytic systems, and AI powered products. And in twenty twenties, we really kind of witnessed what we can call AI democratization. Because with the release of chat GPT in November 2022, generative AI became something that everybody could use. It was no longer something that was a conversation in the data science community or big tech companies that were leveraging machine learning and AI for products and predictive analytics and other tools. It became something that everybody could use. And the reason for that is many. Both the product and the model became much better.

But, also, with the release of Chargebee team in November 2022, we got an amazing user interface that anybody can use. And lang natural language is something that is very natural to humans. And the fact that we could now interact with computing system just by prompts, by language, by text, by telling the system what we want and getting back a response, that's what creating the the the recent hype. But it's not just the hype. It's a true opportunity for every business and personal life, productivity enhancement, and business opportunities. And AI is changing everything we do. The way we do it, the way we work, the way business operate, and it has impact on every industry. And 2025 is definitely the year of Adjantic AI and using AI for coding, using AI agents, and leveraging across every industry and every business.

With big opportunities comes big responsibility, and we have a lot of ethical considerations as we're building all these amazing AI systems. Just to highlight some of the big advancements in AI over the years in the 20 pens, we started seeing AI systems winning over humans, such as in IBM Watson winning Jeopardy. Or in '26 fifteen, sixteen, seventeen, we started seeing AlphaGo that, won over the World Go champion, which is a very strategic game. Alpha zero, the training itself, just by the rules of the games and search libraries and could win over the the time strongest, chess software and started winning chess champion level games. Those were generated by by Google's DeepMind. And then in the '16 and seventeenth, we also had the invention of the transformers architecture that gave the rise to generative AI into the current AI era and really leveraged natural language processing NLP, which is the base for the large language models and and the generative AI foundation.

In 2017, Google researchers group released the paper attention is all you need about the transformers and hugging face BERT bidirectional encoder representations, started generating text, summarizing text, and being used broadly by companies to train those, natural language model systems for deriving sentiments, for answering questions, for summarizing text.

So, all that gave rise to when in 2018, OpenAI released the first GPT-one that was generating a big conversation in the data science community and seemed like the thing, but was not yet there. In the twenty twenties, we saw the future releases of Chargept. Chargept three was already an amazing model that generated a lot of hope. And then, of course, in November 2022, we saw the release of Chargept that started the current AI era hype. It's very important to notice that in 2021, we saw the release of alpha fold, which was a breakthrough in protein folding again by Google's DeepMind. And in 2024, we saw the release of alpha fold three. In at the end of 2024, we finally saw some AI Nobel Prizes for AI. Jeffrey Hinton, who was, one of the pioneers of deep learning, co won the AI Nobel Prize in Physics, and Demis Hassabis co won the Nobel Prize in Chemistry for the invention of AlphaFold.

And in 2023, 2024, 2025, we see the rise of generative AI, large language model, Adjenti KI, that kind of conquer the the world, both in personal use, personal productivity, every business, every use case, and increased adoption. And we're just gonna see more of that because we in a pivotal point in time where AI is gonna change the way we live and work, like many previous revolutions did, And it has a lot of promise and things that we need to take care about such as making sure that this AI is ethical, non biased, that, we develop AI that is ethical and following regulations.

But while we're taking care of all these considerations, there is a huge opportunity with AI across multiple areas, including just enhancing people lives with personalized medicine. So we established AI is all about cats. Just to illustrate how AI models work, we train the model by showing it a lot of examples. For example, if you want the model to identify whether something is a cat or not, we would show it a lot of different examples of cats with different backgrounds because deep learning, we don't know what picks up out from the example. And then we're showing it something that it hadn't seen before, and we ask it to inference. Is it a cat? You can comment in the chat whether you think this is a cat. Well, it is a cat. Well and is this a cat? Please answer in the comments if you think this is a cat. Well, this might be a cat, but maybe it's not a cat. It depends on the context.

AI does not have context. We, human, need to give it context. So when we're thinking of developing AI strategy for our business and scaling AI across our business, we need to do it in the business context. If we're trying to build a business for adopting pets, then it's not a cat. But if we're trying to build a business of classifying felines, then maybe it is a cat. Well, it is a cat because it's a large cat and it's a feline. So it's all about the business context, and we need to put the business context in. There are multiple opportunities for AI across three main verticals, product innovation, process improvement and automation, communication that includes also training and insights.

But we, humans, need to put the context whether we're developing a diagnosis assist tools or drug development AI power tool or personalized medicine empowered by AI, or whether we're kind of doing software development with wipe coding, or we're doing bad fixes using AI, we need to put it in the context of our business needs of what we're trying to build and send more communication, which has a lot of opportunity in customer support currently, but also in marketing, collateral content, documentation.

All that should be used as AI assisting human who is putting the strategy and the content and the context and the thought leadership and using AI as a very capable colleague or assistant or a brainstorming, or brainstorming application or diagnosis assist tool or implementation assist tool.

Again, I keep saying assist. I keep saying colleague because you, the human, driving it. So as we saw, there are a lot of opportunities for AI, but still 80% of AI initiatives fail. And this is because of three main components, the tech issues, the organizational issues, silos, and the adoption issues, the lack of trust, the lack of adoption. But all of these can can be overcome with AI literate leadership. Some of the common challenges are data issues, such as poor data quality, bottlenecks. Your AI can be as good as your data and as good as the human who leading the AI initiative. Talent shortage. You have to have capable engineers and data scientists to build and scale proper initiatives. Stakeholder spine aligning to business objective.

And we can summarize it as lack of overall business wide strategy. So where do you start? At Alta KI, we usually start organizations. We're doing an AI audit, getting prepared, understanding what is the business objectives, how AI can uniquely help, and identifying the gaps. And by using this, we develop an AI strategy across the verticals of product, process, and communication to help scale AI across the in organization and help organization craft a phased implementation roadmap that is both long term planning and can be executed in an agile manner with quickly pivoting based on needs and feedback and the developments in the AI arena.

So developing an AI strategy and road map should start with educating and engaging your stakeholders because AI literacy is key across the whole organization. Transparency brings trust. Ping pointing opportunities, how AI can uniquely help your business and how it aligns with your strategic vision. You have to make sure that whichever AI initiatives you wanna scale are closely aligned with your overall business objectives. And you need to upscale and allocate talent to implement those. You need to communicate widely to your whole organization. You should start with small pilots when you launch launching your AI throughout their organization, starting with small pilots, starting with experimentation, ensuring data quality governance, and compliance with regulations is key from an early stage. And after the experimentation and developing prototypes, you need to move from this siloed personal productivity gains by moving from tactical to the strategic to scale across your whole organization.

You need to scale responsibly and take it from the siloed experimentation and prototyping to having an, an organizational wide strategy that is unified aligned within the all stakeholders has a lot of transparency and a lot of AI literacies through the through the implementation and implementing it with the phase stages of the road map, starting with a small pilot, low hanging fruits, the quick successes, while ensuring the ethical and safety in implementation by conducting thorough testing and pivoting quickly as required throughout the implementation because the development are out there, and the feedback and development should allow you to pivot quickly through an agile implementation road map.