Free AI Threatens Giants News | AviaryAI Newsletter
The open source AI community has delivered a landmark result: freely available models now match or surpass GPT-4 performance on key benchmarks. For credit union technology leaders, this development changes the AI vendor landscape significantly -- not because credit unions should rush to deploy unvetted open source models, but because competitive pressure from free alternatives will drive down the cost of enterprise AI and give institutions more leverage in vendor negotiations.
Silicon Valley Stunned: Open Source Models Threaten Giants’ Dominance
Meta's Llama 3.1 and Mistral's Large 2, two open-source AI models, have reportedly achieved "frontier-level" status, matching or surpassing proprietary systems from industry leaders like OpenAI and Google. This sudden democratization of cutting-edge AI capabilities could accelerate innovation globally, reshape industries, and alter the balance of power in the tech world.
So what?
As powerful AI becomes freely available, expect a surge of innovative, cost-effective fintech solutions leveraging these models. Credit union executives should prepare to re-evaluate existing tech partnerships and be open to nimble startups that may now offer cutting-edge AI capabilities at a fraction of the cost of established vendors.
OpenAI Unveils SearchGPT: A New AI-Powered Search Engine
OpenAI has introduced SearchGPT, an AI-powered search engine that organizes and summarizes web information in response to user queries. Initially available to 10,000 test users, SearchGPT aims to provide more contextualized search results than traditional engines, with plans to eventually integrate the feature into ChatGPT.
So what?
This represents a shift in how people will access and process information, potentially impacting how credit union members research and discover financial products and services. Credit unions should prepare for new member acquisition strategies that incorporate these channels of discovery.
AI Scores Silver at Math Olympics: A New Era of Problem-Solving
DeepMind's AI systems, AlphaProof and AlphaGeometry 2, achieved a silver medal standard at the International Mathematical Olympiad, solving four out of six problems. This breakthrough demonstrates AI's growing capacity for advanced mathematical reasoning, potentially unlocking new frontiers in science and technology.
So what?
This news represents a leap in AI's ability to tackle complex, multi-step reasoning tasks. This could soon translate to AI systems that can navigate intricate regulatory landscapes, recommend personalized products or services, and detect sophisticated fraud patterns.
Read the full story here.
THE AI EXECUTIVE'S HANDBOOK
One simplified GenAI concept per week to build your AI Acumen
Citi’s AI Report: Threat or Opportunity?
Citi's latest report on AI in banking has been making waves, with headlines proclaiming "Half of Banking Jobs at Risk of AI Automation." But what does this really mean for credit unions? Let's dive into the data and explore the implications.
Key Findings from Citi's Report:
- 54% of banking "work-time" has high automation potential
- 12% of work-time has high augmentation potential
- 93% of financial institutions expect AI to boost profits within 5 years
- AI could add $170 billion to global banking profits by 2028
Our take: AI won’t take your job. The person who knows how to use it will.
And that’s good news for credit unions… If they can seize the opportunity:
If 1 person can do the work of 2, CU’s can compete more effectively with larger institutions. But a “wait and see” approach will give larger institutions more time to capture the upside.
Automation of routine tasks frees up staff to focus on high-value, relationship-building activities. The human touch that credit unions are known for can be amplified, not replaced.
As non-profits, cost savings from automation can be reinvested directly into member benefits or new services
If anything, Citi’s report is a(nother) wake-up call. Credit unions that act now to integrate AI into their operations and culture will thrive in this new landscape, maintaining their unique value proposition while competing effectively with larger institutions.
Foundational Models
Foundational Models are large-scale AI systems trained on vast amounts of diverse data, which serve as a base for creating more specialized AI applications. These models can then be adapted to perform more specific tasks without needing to be completely retrained from scratch. They're called "foundational" because they provide a strong base of knowledge and capabilities that can be built upon for many different AI applications.
Google’s AI Spending Spree
Alphabet, Google's parent company, reported strong Q2 2024 earnings, beating analyst expectations with $84.7 billion in revenue and $1.89 earnings per share. While the company showed growth across various segments, a key highlight was the significant increase in AI investments. Alphabet spent $2.2 billion on AI model development in Q2, doubling the amount from the previous year.
Despite these massive investments, the timeline for realizing substantial AI-driven revenue remains uncertain. Analysts suggest that material AI revenue is more likely to materialize in 2025-2026, as most companies are still in the pilot phase of AI adoption. This indicates a significant gap between current AI expenditures and future returns, highlighting the speculative nature of these investments.
So what?
While big tech companies are pouring billions into AI without immediate returns, many businesses outside the tech sector are already seeing tangible benefits from targeted AI implementations.
The EY survey of business leaders (including the finance sector), we covered last week showed:
95% of respondents are currently investing in AI, with many planning to at least double their investment by 2025.
56% are building their own solutions while the other half are buying ready-made AI products
77% reported improvements in operational efficiency
75% reported improved employee productivity
72% reported increased customer satisfaction
This suggests that the real value of AI in the near-term lies not in moonshot projects, but in practical, focused applications.
Frequently Asked Questions
Should credit unions use open source AI models instead of paid platforms?
Open source models offer flexibility and cost savings but require technical resources to deploy, fine-tune, and maintain -- capabilities most credit unions don't have in-house. Managed platforms that combine open source infrastructure with financial services specialization typically offer better value.
How does open source AI competition affect the pricing of platforms like AviaryAI?
Competitive pressure from open source alternatives keeps enterprise AI pricing in check and incentivizes commercial providers to deliver clear value-adds: financial domain training, compliance features, integration support, and ongoing model maintenance that open source cannot provide out of the box.
What AI math capabilities are useful for financial services?
Advanced AI math capabilities enable better loan amortization analysis, risk scoring explanations, budget scenario modeling, and fraud anomaly detection. These capabilities are increasingly available as native features in AI platforms targeting financial services.
