Are Small Language Models the Future of Affordable (and Better) AI?
[Creatives on Call]

Are Small Language Models the Future of Affordable (and Better) AI?

Today’s question: Want an AI language model that gives more accurate results and saves you money?

When it comes to generative AI and the models that drive them, sometimes less is more. Small language models, trained on much smaller but very specific, high-quality datasets, are far less expensive to train and maintain, and often outperform the kitchen-sink approach of gigantic, multipurpose large language models. 

Here we explain how these mini marvels, geared toward a single task, deliver the power of generative AI for a fraction of the cost. 


[Slack]

AI is the cure for burnout, but…

AI usage is up 23% since the beginning of the year, according to Slack's new Workforce Index, but employees aren’t allocating the time saved by AI toward strategic or high-value activities like learning and skill building. Instead, they’re likely to spend 37% more of their time on routine administrative tasks.



What our community is saying

“Harnessing AI to transform customer data into personalized marketing and commerce experiences is a game-changer. The built-in Einstein Trust Layer adds an essential element of security and reliability, making the CRM a standout choice for businesses looking to innovate and build deeper customer relationships.” 

-Nikhil Khetan, CEO and founder, Techabled Consulting - Authorized Salesforce Reseller Partner


What we’re reading

  • The cost of training the world’s largest AI models is growing by a factor of two to three per year, putting it on pace to top $1 billion by 2027. Two drivers: computational power and compensation. (Epoch AI

  • To cut down on returns, Amazon is turning to generative AI and computer vision to detect defective, incorrect, or damaged products before shipping. (Fast Company)

  • Sixty-five percent of organizations polled by McKinsey are regularly using generative AI, twice as many as just ten months ago. They’re already seeing significant cost savings and increased revenue. (McKinsey)


Dreamforce Is Coming – Register by June 27 for 20% Off

Join us for three days of product innovations across CRM AI Data Trust, and experience the biggest shift in technology of our time. Learn how other customers are revolutionizing their business with AI and be inspired to design your AI Enterprise roadmap. 

Use code DF24EARLY to save over $1,200. But register now, this offer expires June 27, 2024. 

[Salesforce]

This newsletter was curated by Lisa DiCarlo Lee, Contributing Editor at Salesforce.


What’s the biggest question you have about AI? Let us know in the comments. We read each one and may just feature yours in a future newsletter!

AI's good performance is only possible with clean data. ;)

Oscar Cabrera

Gerente de operaciones y CEO | Planificación y conciencia ambiental I 1M Followers

3w

Joining!

Lambo Zou

Your China Sourcing Agent - Furniture Hardware Expert | Smart Drawer Lock | Cabinet Handle | Drawer Slides | Furniture Leg | Hinge | Cable Grommet

3w

AI is the trends and future.

Max Greene

Senior Customer Success Manager - Empowering organizations to improve customer and employee experiences with Artificial Intelligence

3w

There is a niche for small language models. Domain expertise is a real thing and leveraging targeted datasets is not just cost effective and efficient but also avoids the risk of hallucination. This also means that humans (yes, I said it), can support the delivery of small language models with their own deep understanding of domain specific use cases.

ShreeKanth Reddy G

Pursuing EA |꧁༺432K Impressions༻꧂| Senior CRM - US Taxation | Passionate & Upcoming Singer & Composer |

3w

Interesting…Small language models might just be the future of affordable AI, and they’re like the compact cars of the tech world. Imagine driving a sleek, efficient smart car instead of a gas-guzzling monster truck. These petite powerhouses can run on less powerful hardware, saving you a bundle on energy bills—your laptop won’t sound like it’s about to take off anymore. They also need less data for training, so even your grandma’s dusty old computer can get in on the AI action. Plus, they can work on edge devices, giving you speedy responses without relying on the cloud—perfect for when you’re hiding from the Wi-Fi in your basement. These small but mighty models could become the budget-friendly heroes of the AI world, proving that sometimes, size really doesn’t matter!

To view or add a comment, sign in

Insights from the community

Others also viewed

Explore topics