Meta AI Claims to Compete with ChatGPT: A Test for Trip Planning
Overview of Meta AI’s Latest Model
Recently, Meta AI launched its latest generative AI model, known as Llama 3.1 405B. According to Meta, this new version surpasses competitor models from OpenAI in almost every measurable aspect. However, our focus is on understanding how this model performs in practical applications, particularly in the travel sector.
Practical Performance Tests
A test conducted by Skift in 2023 on the previous version of Meta’s chatbot showed some limitations, particularly compared to its competitors. With the new version now available through WhatsApp and on the Meta.ai website, we decided to evaluate its current performance.
Accessibility and Features
Unlike OpenAI’s ChatGPT, which has robust capabilities like image reading for menu translations and audio listening for speech translations, Meta AI’s chatbot does not yet support these features. Additionally, it lacks a reliable connection to flight information, unlike Google’s Gemini, which offers direct access to flight details.
Meta AI has taken a different approach by making their model “open,” allowing developers to download and modify the code. This is in contrast to closed models like ChatGPT, Gemini, and Anthropic’s Claude.
Things to Do in Vancouver
To test the chatbot, we prompted it to suggest activities in Vancouver focusing on hiking and cultural experiences. The suggestions were organized by type and included recommendations for particular days, making it easier to plan.
For instance:
- Cypress Mountain and Stanley Park: Suggested for the same day due to their proximity.
- Garibaldi Lake: Recommended as a single-day trip due to its distance from Vancouver.
Unlike traditional trip planners, Meta AI seemed to consider logistical aspects effectively, offering relevant suggestions without straying too far from the city.
Finding Restaurants
The chatbot provided some interesting restaurant options in Vancouver. However, two of the recommendations, West Restaurant and Heirloom, are no longer in operation. This highlights a potential issue: users need to verify the existence of these restaurants after receiving recommendations, as the chatbot didn’t provide links or reliable web access to check these establishments.
Searching for Flights
When we asked the chatbot to find a flight from New York City to Vancouver for a specific date, it struggled. It could not provide direct flight options but instead offered a link to a Bing search for flights. This step indicates that for precise flight details, users may find using a traditional search engine more effective.
Searching for Hotels
The hotel search functionality proved to be even less helpful than the flight search. When prompted to find accommodations near downtown Vancouver, the chatbot gave names but provided inaccurate pricing and no links to book these properties. It instead suggested a link to a Google search for hotels, again making traditional searches seem more practical.
Helpful Packing List
One feature offered by the Meta AI chatbot is a packing list generator. When prompted for suggestions based on our Vancouver trip—which would last a week and include both hiking and cultural experiences—the chatbot generated a reasonable list tailored to the nature of the trip. While many items were common sense, this feature can be particularly helpful for travelers who might overlook essentials.
Summary of Chatbot Capabilities
While the new Llama 3.1 version has shown improvements in suggesting travel activities, it still has limitations in flight and hotel searches. The chatbot can create useful packing lists but lacks vital features like real-time updates and refinement of its restaurant recommendations. As Meta continues to develop this technology, it will be essential to monitor how effectively it meets users’ evolving needs in the travel sector.
