AI is not a new field, and Xelleron’s AI experts have extensive experience, predating the recent surge of AI providers. This enables us to collaborate with clients to implement AI solutions in an effective and cost-efficient manner.

Given the powerful nature of AI, it can be tempting to use it everywhere. However, it’s essential to choose the right approach, considering factors like accuracy, token costs, model selection, data privacy, confidentiality, intellectual property, and compliance with regulatory changes, such as those in the EU.

With new AI models being released frequently, it’s crucial to design AI software solutions in a flexible manner to incorporate future advancements.

Case Study

A client needs to extract key text from a filled-out pdf form that contains private information. To maintain data privacy, we can use a traditional parsing algorithms to separate the private data before passing the remaining information to a GPT for further processing. This approach not only protects sensitive data but also reduces token usage and cost.

Historical Perspective

Over time, AI has matured, and with the rise of large language models like ChatGPT, we have entered an era of widely accessible AI.

• 1950: Alan Turing introduced the Turing Test, setting the foundation for AI.
• 1956: The Dartmouth Conference coined the term “Artificial Intelligence.”
• 1957: Frank Rosenblatt developed the Perceptron, one of the first neural network algorithms.
• November 2022: ChatGPT-3.5 was released by OpenAI. This version introduced several improvements in language understanding, response coherence, and overall performance compared to its predecessors.
• October 2024: There are numerous major AI models, and the number keeps growing as technology advances. The major categories include:
  • Large Language Models (LLMs): GPT-3, GPT-4, BERT, T5.
  • Vision Models: Vision Transformers (ViT), ResNet, EfficientNet.
  • Generative Models: DALL-E, StyleGAN, VQ-VAE.
  • Reinforcement Learning Models: AlphaZero, Deep Q-Networks (DQN).
  • Multi-modal Models: CLIP, Flamingo.
  • Conversational Agents: ChatGPT, LaMDA, Bard.

And on top of that, each of these categories has several prominent models and variants, making it hard to define an exact number of “major” models.

At Xelleron, AI has dramatically increased our developers’ productivity, transforming them into “prompt engineers” who harness the power of AI for software development.