Industry experts are debating the business impact of Meta’s free artificial intelligence (AI) model Llama 3.1, weighing its potential against practical implementation challenges.
Boasting 405 billion parameters, the AI model claims performance comparable to proprietary competitors like GPT-4 and Claude 3.5 Sonnet. As Meta expands its reach, with CEO Mark Zuckerberg predicting it will become the most widely used AI assistant by year’s end, businesses are weighing the implications of access to powerful, cost-free AI against the challenges of implementation and security.
“These models can be used to communicate with customers and provide instant 24/7 assistance with simple queries that do not require human intervention,” Ilia Badeev, head of data science at Trevolution Group, told PYMNTS. “With LLMs [large language models], marketing campaigns and recommendations can be truly personalized for individual customers.”
Some experts predict a fundamental shift in customer service. “If you think about the cost of intelligence effectively going to zero over time for customer relations, call centers will not exist in the future. AI systems will manage huge volumes of customer inbound in a meaningful and satisfactory way to the end user,” Mike Conover, CEO of the AI company Brightwave, told PYMNTS.
The potential for businesses to customize these models is significant. “By fine-tuning Llama on their specific domain data, companies can create powerful natural language interfaces that understand customer queries, provide intelligent recommendations, and automate tasks like product categorization and content generation,” Hamza Tahir, CTO and co-founder of ZenML, an open-source machine learning operations (MLOps) startup, told PYMNTS.
The availability of powerful, open-source AI models could level the playing field for smaller businesses. “Open-source models like Llama have the potential to democratize AI-powered commerce tools for small businesses and startups,” Tahir said.
“Even small teams can leverage state-of-the-art natural language processing capabilities to build intelligent chatbots, product recommenders and content generators.”
Open-source AI also offers advantages in regulatory compliance. “Processing data with in-house models keeps user data private and compliant with regulatory laws (such as GDPR),” Badeev pointed out, referring to the EU’s General Data Protection Regulation. This contrasts with proprietary models that may require sending user data to third-party services.
The introduction of Llama 3.1 is stirring debate about its potential to disrupt the commercial AI market. Conover said that the 405 billion-parameter model from Llama is comparable in its reasoning quality to OpenAI’s GPT-4. “This means commercial providers do not have some secret sauce that would lead to vendor lock-in — business owners are the masters of their own destiny,” he added.
Tahir predicted that the introduction of the new model may presage a shift toward a service-based model, where AI companies differentiate themselves through their domain expertise, data assets and ability to customize and deploy open-source models for specific use cases.
The economic impact could be substantial. “For business owners like eCommerce platforms and customer service providers, you’re going to see improving unit economics for these services because of the competitive pressures that open-source technologies place on the commercial providers,” Conover added.
Despite the opportunities, businesses face challenges in implementing open-source AI. “Open-source AI models give SMEs [small to mid-sized enterprises] the advantage of doing more and reaching a wider audience, but this comes at the cost of both talent and security, Harry Toor, chief of staff at OpenSSF, which promotes open-source software, told PYMNTS.
He added, “Open-source AI models need to be consumed securely to ensure outputs aren’t manipulated, which could sink any SME operating in a cost-constrained environment.”
Security measures are crucial. “Secure open-source AI models should be built from a secured development environment, cryptographically signed, and follow best practices already in place for open-source software development. This can be achieved by leveraging existing open-source tools from OpenSSF and elsewhere to secure open-source AI models,” Toor said.
Potential supply chain issues also pose a risk. “The commercial AI market needs to evaluate the supply chain for open-source AI models. Recent global cyber issues like XZ Utils and the Microsoft Blue Screen of Death have shown that widely used software components can cripple industries,” Toor warned.
As businesses consider adopting open-source AI, they face a complex set of considerations. The long-term implications for eCommerce, customer service and marketing strategies are still unfolding. While some predict a radical transformation of these sectors, others caution that the technology’s impact will depend on factors beyond mere availability.
Open-source models could lead to more effective feedback collection. “User feedback/reactions can be effectively gathered from various sources such as reviews, social media mentions, and customer support interactions. These can be massively processed using AI to extract insights and analytics instantly,” Badeev noted.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
