Meta AI recently released a new version of its generative AI model called Llama 3.1 405B, claiming it outperforms OpenAI in many metrics. Skift conducted a test to explore the performance of this model in travel-related tasks. The chatbot based on this new model showed some improvements but still fell short when compared to competitors. While the model is open and accessible in the U.S. via WhatsApp and the Meta.ai website, it lacks some features such as image and audio recognition. Meta AI also does not have a reliable connection to flight information, unlike Gemini by Google.
In a test focusing on activities like hiking and cultural experiences in Vancouver, the chatbot suggested activities based on the day needed to complete them. The chatbot seems to be getting better at taking logistics into account, suggesting activities that make sense considering the proximity to each other. However, the chatbot did not suggest visits too far outside Vancouver. When asked for restaurant suggestions, the chatbot provided some solid recommendations but failed to provide links to their websites. This means users would need to complete a traditional search to verify the restaurant’s existence and get more details.
The flight search feature on the chatbot did not work well, failing to find options for a flight from New York City to Vancouver in September. While it did provide a link to a Bing search showing options for that time period, it may be more efficient to start the flight search in a traditional search engine. Similarly, the hotel search feature was also not useful, providing inaccurate pricing information for hotels near downtown Vancouver without links to the properties. The chatbot did provide a link to a Google search for hotel tonight, which was not helpful for the specific request.
One interesting feature of Meta AI’s chatbot is its ability to provide a packing list for an upcoming trip. By providing the destination, number of days, and type of trip, the chatbot can suggest items to pack based on the trip’s nature. While most of the suggestions are common sense, it could be helpful for those who tend to forget items when packing. Overall, Skift found that Meta AI’s AI model has shown some improvements in trip planning capabilities compared to previous versions but still lags behind competitors in terms of features and reliability. Users may still find traditional search methods more effective for planning flights, hotels, and other travel arrangements.
