Yandong Liu is the Co-founder and CTO at Connectly.ai. He previously worked at Strava as a CTO. Yandong Liu attended Carnegie Mellon University.
Founded in 2021, Connectly is the leader in conversational artificial intelligence (AI). Using proprietary AI models, Connectly’s platform automates how businesses communicate with their customers and sell their products across any messaging platform. Connectly enables your entire customer journey – from sales and marketing to customer experience and support – to be conducted throughout the customer’s preferred messaging platform.
Are you able to share the genesis story behind Connectly?
Connectly was born from the vision of becoming the leader in conversational AI. My co-founder, Stefanos, and I met through a mutual friend within the founder community and bonded over a shared passion for the longer term of messaging. With my background leading technology teams at Strava and Uber and Stefanos’s experience overseeing Facebook Messenger, we got down to create the AI-powered infrastructure of the longer term, helping businesses take advantage of their customer messages in an increasingly complex ecosystem.
What exactly are Small Language Models (SLMs), and the way do they differ from Large Language Models (LLMs)?
SLMs are AI models designed to grasp and generate human language but with fewer parameters and computational requirements in comparison with Large Language Models. Within the context of AI marketing solutions for messaging platforms like WhatsApp and Instagram, SLMs provide faster response times and could be easily deployed on quite a lot of devices, making them ideal for real-time customer interactions. Their smaller size allows for efficient performance without compromising the standard of responses.
Are you able to discuss how SLMs reduce the likelihood of hallucinations and improve the reliability of AI responses?
SLMs reduce the likelihood of hallucinations—instances where AI generates incorrect or nonsensical information—by specializing in a smaller, more manageable set of parameters. For AI-messaging based marketing solutions, this focused approach ensures more predictable and reliable responses, enhancing customer trust and engagement. The reduced complexity of SLMs minimizes the possibilities of generating off-topic or erroneous content, thereby improving the general reliability of AI interactions.
Are you able to explain why SLMs are particularly useful for retailers, especially within the context of chatbots?
On account of the big amounts of information LLMs are fed with, they are sometimes slow. Nonetheless, messaging and conversational commerce require a faster response time with the intention to higher and more accurately serve customers. For retailers, SLMs are more practical and useful attributable to the extent of detail they will provide within the retail industry. Moreover, SLMs are sometimes cheaper because they’re more agile, meaning every retail company, from a small startup to an enormous online retailer, can utilize them.
How do SLMs offer more personalized experiences for patrons in comparison with LLMs?
SLMs offer more personalized experiences for patrons by being easier to fine-tune for specific tasks and domains. Their smaller size allows for quicker and more efficient customization, enabling businesses to tailor the models to the unique needs and preferences of their customers. This focused customization leads to more relevant and personalized interactions, enhancing the shopper experience.
How does Connectly integrate SLMs into its platform to reinforce e-commerce capabilities?
We integrate SLMs into our platform to reinforce e-commerce capabilities by leveraging their efficiency and adaptableness. These models enable quick and accurate customer interactions on messaging platforms like WhatsApp and Instagram, providing personalized product recommendations and easy customer support. The lightweight nature of SLMs ensures that responses are fast and relevant, improving the general customer experience and driving engagement.
What are some specific examples of how retailers have successfully implemented SLMs of their operations?
Our clients are having great success with SLMs. A fashion retailer is using SLMs to supply personalized styling advice through WhatsApp, recommending outfits based on the shopper’s previous purchases and preferences. Similarly, an electronics retailer deployed SLMs on Instagram to reply customer queries about product features and availability in real time, enhancing the shopping experience and reducing the load on customer support teams.
Why should retailers consider transitioning from LLMs to SLMs for his or her specific business applications?
Retailers should consider transitioning from LLMs to SLMs for his or her specific business applications attributable to the increased efficiency and cost-effectiveness of SLMs. SLMs are faster, require less computational power, and could be easily fine-tuned for specific tasks, making them ideal for real-time customer interactions on messaging platforms like WhatsApp and Instagram. This transition can result in more responsive and personalized customer support while reducing operational costs.
What future advancements in SLM technology are you most enthusiastic about?
I’m most enthusiastic about advancements in SLM technology that may further enhance their efficiency and accuracy. As an illustration, improvements in transfer learning and fine-tuning techniques will allow SLMs to change into even more proficient at specific tasks with minimal data. Moreover, the mixing of SLMs with multimodal capabilities—combining text, voice, and image data—will enable richer and more interactive customer experiences on platforms like WhatsApp and Instagram. These advancements will make SLMs much more helpful for retailers trying to provide personalized and interesting customer interactions.
How do you see the adoption of SLMs evolving in the following few years throughout the retail industry?
I see the adoption of SLMs within the retail industry growing significantly. As retailers proceed to hunt more efficient and cost-effective ways to have interaction with customers, the speed and adaptableness of SLMs will change into increasingly helpful. SLMs will probably be integrated more widely into customer support platforms, marketing campaigns, and personalized shopping experiences on messaging apps like WhatsApp and Instagram, even on TikTok. This shift will help retailers provide quicker, more personalized interactions, enhancing customer satisfaction and loyalty.
