AI Connect Newsletter | Edition #12
Welcome to the 12th edition of the Prescience AI Connect Newsletter! Each week, AI Connect brings you the latest updates, trends, insights, expert opinions, and real-life stories from the world of AI and data analytics.?
As we draw the close of the year,?we decided to look?at the mirror view and skim through some key announcements from leading LLM models in 2024, along with selective tech announcements?from notable tech giants. You will also find valuable insights from McKinsey on the evolving landscape of AI.?
And before we let you go, don’t forget to check out the latest episode of the ‘AI Connect’ podcast, where Anirban Majumder and Jaydip Sikdar discuss the top Data and AI Predictions for 2025!?
Overview of Key LLM Models
Open AI - GPT-4o and its evolution?
GPT-4o ("Omni") is a multimodal AI model that analyzes and generates text, images, and sound, offering twice the speed and half the cost of GPT-4 Turbo. It's free within usage limits, with extended access for ChatGPT Plus subscribers. GPT-4o Mini, a smaller, cost-effective version, replaced GPT-3.5 Turbo in ChatGPT. The enhanced GPT-o1 model succeeded it, offering superior reasoning for complex tasks like programming and scientific applications.
Meta?
Llama 3, available in 8B and 70B sizes, was trained on 15 trillion tokens, outperforming models like Gemini Pro 1.5 and Claude 3 Sonnet. Llama 3.1 expanded to 8B, 70B, and 405B sizes, enhancing its capabilities. Llama 3.2 improved speed and accuracy across tasks. Llama 3.3 is set to advance multi-modal capabilities, integrating visual input with text to push AI performance further.
Anthropic
Introduced in June 2024, Claude 3.5, also known as Claude 3.5 Sonnet, emphasized visual reasoning and cost efficiency, contributing to Anthropic's focus on AI safety and public benefit.
Microsoft?
Microsoft's Phi-3, launched in April 2024, and Phi-4, introduced in December 2024, are small language models with 14 billion parameters, trained on 4.8 trillion tokens, and available under the MIT license, reflecting Microsoft's commitment to advancing AI accessibility.
Google DeepMind
Google DeepMind's Gemini 1.5, released in February 2024, introduced a one million-token context window, enhancing performance in complex tasks. Later in 2024, Gemini 2.0 further refined these capabilities, improving efficiency, handling larger datasets, and providing more accurate and coherent responses. Both models push the limits of AI language modeling.
Read more.??
Emerging Technology Developments??
NVIDIA??
NVIDIA unveiled the Blackwell AI chip for enhanced processing in data centers and AI apps. They also introduced NVLM 1.0, an open-source multimodal model with 72 billion parameters, aiming to compete with GPT-4.??
Apple??
Apple launched Apple Intelligence, an AI-driven suite enhancing Siri and other system functions. This integration offers more natural interactions, including the ability to type queries instead of speaking them, and provides intelligent assistance within the Photos app.??
Salesforce?
Recently, Salesforce unveiled Agentforce 2.0, an advanced version of its AI-powered agent program. This update enhances integration with Slack and offers improved reasoning, integration, and customization capabilities. The full release is anticipated in February 2025.??
?
Trends & Insights??
McKinsey's report predicts AI software and services could generate $15.5 trillion to $22.9 trillion by 2040, with generative AI contributing $2.6 trillion to $4.4 trillion in enterprise applications like customer operations and marketing.?
Quote of the week??
"AI has the potential to revolutionize productivity, but if its benefits are hoarded by the wealthy, it could sow the seeds of inequality, paving the way for the rise of fascism."?– Geoffrey Hinton?
Prescience Tidbits?
In the latest episode of the ‘AI Connect' podcast, Anirban Majumder and Jaydip Sikdar met up to discuss 2025 AI and data analytics predictions. ?
Click here to view the full podcast.??
About Prescience
Prescience is an AI and data analytics company that works with large Fortune 500 companies and midsized businesses globally, helping them unlock the power of enterprise data, thus enabling faster and better business decision-making.
Founded in 2017, the company offers comprehensive data solutions and services that integrate AI and machine learning across analytics, business intelligence, data engineering, and more—driving measurable business value and ROI for its customers.
You can read some of the customer success stories here.
We want to hear from you about the AI Connect Newsletter. Please share your thoughts, feedback, and ideas in the comments section.?