nlp-mistral-prompting.ipynb

{"metadata":{"jupytext":{"formats":"ipynb,py:light"},"kernelspec":{"name":"python3","display_name":"Python 3","language":"python"},"language_info":{"name":"python","version":"3.10.13","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat_minor":5,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# Prompting Capabilities \n\n- Note, you can try any of these prompts outside of this classroom, and without coding, by going to the chat interface [Le Chat](https://chat.mistral.ai/chat).\n  - You can sign up with a free account.\n  - Signing up for an account is **not** required to complete this course.","metadata":{}},{"cell_type":"code","source":"!pip install mistralai","metadata":{"height":30,"execution":{"iopub.status.busy":"2024-05-04T08:05:44.034198Z","iopub.execute_input":"2024-05-04T08:05:44.034826Z","iopub.status.idle":"2024-05-04T08:06:01.991361Z","shell.execute_reply.started":"2024-05-04T08:05:44.034771Z","shell.execute_reply":"2024-05-04T08:06:01.989750Z"},"trusted":true},"execution_count":1,"outputs":[{"name":"stdout","text":"Collecting mistralai\n  Downloading mistralai-0.1.8-py3-none-any.whl.metadata (1.8 kB)\nCollecting httpx<0.26.0,>=0.25.2 (from mistralai)\n  Downloading httpx-0.25.2-py3-none-any.whl.metadata (6.9 kB)\nRequirement already satisfied: orjson<4.0.0,>=3.9.10 in /opt/conda/lib/python3.10/site-packages (from mistralai) (3.9.10)\nRequirement already satisfied: pydantic<3.0.0,>=2.5.2 in /opt/conda/lib/python3.10/site-packages (from mistralai) (2.5.3)\nRequirement already satisfied: anyio in /opt/conda/lib/python3.10/site-packages (from httpx<0.26.0,>=0.25.2->mistralai) (4.2.0)\nRequirement already satisfied: certifi in /opt/conda/lib/python3.10/site-packages (from httpx<0.26.0,>=0.25.2->mistralai) (2024.2.2)\nRequirement already satisfied: httpcore==1.* in /opt/conda/lib/python3.10/site-packages (from httpx<0.26.0,>=0.25.2->mistralai) (1.0.5)\nRequirement already satisfied: idna in /opt/conda/lib/python3.10/site-packages (from httpx<0.26.0,>=0.25.2->mistralai) (3.6)\nRequirement already satisfied: sniffio in /opt/conda/lib/python3.10/site-packages (from httpx<0.26.0,>=0.25.2->mistralai) (1.3.0)\nRequirement already satisfied: h11<0.15,>=0.13 in /opt/conda/lib/python3.10/site-packages (from httpcore==1.*->httpx<0.26.0,>=0.25.2->mistralai) (0.14.0)\nRequirement already satisfied: annotated-types>=0.4.0 in /opt/conda/lib/python3.10/site-packages (from pydantic<3.0.0,>=2.5.2->mistralai) (0.6.0)\nRequirement already satisfied: pydantic-core==2.14.6 in /opt/conda/lib/python3.10/site-packages (from pydantic<3.0.0,>=2.5.2->mistralai) (2.14.6)\nRequirement already satisfied: typing-extensions>=4.6.1 in /opt/conda/lib/python3.10/site-packages (from pydantic<3.0.0,>=2.5.2->mistralai) (4.9.0)\nRequirement already satisfied: exceptiongroup>=1.0.2 in /opt/conda/lib/python3.10/site-packages (from anyio->httpx<0.26.0,>=0.25.2->mistralai) (1.2.0)\nDownloading mistralai-0.1.8-py3-none-any.whl (15 kB)\nDownloading httpx-0.25.2-py3-none-any.whl (74 kB)\n\u001b[2K   \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m75.0/75.0 kB\u001b[0m \u001b[31m2.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n\u001b[?25hInstalling collected packages: httpx, mistralai\n  Attempting uninstall: httpx\n    Found existing installation: httpx 0.27.0\n    Uninstalling httpx-0.27.0:\n      Successfully uninstalled httpx-0.27.0\n\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\njupyterlab 4.1.6 requires jupyter-lsp>=2.0.0, but you have jupyter-lsp 1.5.1 which is incompatible.\njupyterlab-lsp 5.1.0 requires jupyter-lsp>=2.0.0, but you have jupyter-lsp 1.5.1 which is incompatible.\u001b[0m\u001b[31m\n\u001b[0mSuccessfully installed httpx-0.25.2 mistralai-0.1.8\n","output_type":"stream"}]},{"cell_type":"code","source":"from kaggle_secrets import UserSecretsClient\nuser_secrets = UserSecretsClient()\nsecret_value_0 = user_secrets.get_secret(\"MISTRAL_KEY\")","metadata":{"execution":{"iopub.status.busy":"2024-05-04T08:07:52.621091Z","iopub.execute_input":"2024-05-04T08:07:52.621552Z","iopub.status.idle":"2024-05-04T08:07:52.805457Z","shell.execute_reply.started":"2024-05-04T08:07:52.621513Z","shell.execute_reply":"2024-05-04T08:07:52.803995Z"},"trusted":true},"execution_count":2,"outputs":[]},{"cell_type":"markdown","source":"- Notice that it's \"mistralai\", and not \"mistral\"","metadata":{}},{"cell_type":"markdown","source":"### Load API key and helper function\n- Note: You can view or download the helper.py file by clicking on the \"Jupyter\" logo to access the file directory.","metadata":{}},{"cell_type":"code","source":"from mistralai.client import MistralClient\nfrom mistralai.models.chat_completion import ChatMessage\nmistral = MistralClient(api_key=secret_value_0)\nmodel = \"open-mixtral-8x7b\"\n\nchat_response = mistral.chat(\n    model=model,\n    messages=[ChatMessage(role=\"user\", content=\"What is the best French cheese?\")]\n)\n\nprint(chat_response.choices[0].message.content)","metadata":{"height":47,"execution":{"iopub.status.busy":"2024-05-04T08:10:10.037752Z","iopub.execute_input":"2024-05-04T08:10:10.038172Z","iopub.status.idle":"2024-05-04T08:10:12.400177Z","shell.execute_reply.started":"2024-05-04T08:10:10.038138Z","shell.execute_reply":"2024-05-04T08:10:12.398901Z"},"trusted":true},"execution_count":10,"outputs":[{"name":"stdout","text":"It's impossible to declare a single \"best\" French cheese, as cheeses vary greatly in taste, texture, and flavor, and preferences for cheese are highly subjective and can depend on the specific context or dish it's being used for.\n\nHowever, some of the most well-known and highly regarded French cheeses include:\n\n* Brie: a soft, creamy cheese with a white, edible rind.\n* Camembert: similar to Brie, but usually stronger in flavor.\n* Roquefort: a blue cheese made from sheep's milk, known for its tangy, robust flavor.\n* Comté: a hard, nutty cheese made from cow's milk.\n* Reblochon: a soft, washed-rind cheese with a strong, earthy flavor.\n\nUltimately, the best French cheese for you will depend on your personal preferences and the specific dish or occasion you're using it for. It's worth trying a variety of different types to find your favorite!\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## Classification","metadata":{}},{"cell_type":"code","source":"prompt = \"\"\"\n    You are a bank customer service bot. \n    Your task is to assess customer intent and categorize customer \n    inquiry after <<<>>> into one of the following predefined categories:\n    \n    card arrival\n    change pin\n    exchange rate\n    country support \n    cancel transfer\n    charge dispute\n    \n    If the text doesn't fit into any of the above categories, \n    classify it as:\n    customer service\n    \n    You will only respond with the predefined category. \n    Do not provide explanations or notes. \n    \n    ###\n    Here are some examples:\n    \n    Inquiry: How do I know if I will get my card, or if it is lost? I am concerned about the delivery process and would like to ensure that I will receive my card as expected. Could you please provide information about the tracking process for my card, or confirm if there are any indicators to identify if the card has been lost during delivery?\n    Category: card arrival\n    Inquiry: I am planning an international trip to Paris and would like to inquire about the current exchange rates for Euros as well as any associated fees for foreign transactions.\n    Category: exchange rate \n    Inquiry: What countries are getting support? I will be traveling and living abroad for an extended period of time, specifically in France and Germany, and would appreciate any information regarding compatibility and functionality in these regions.\n    Category: country support\n    Inquiry: Can I get help starting my computer? I am having difficulty starting my computer, and would appreciate your expertise in helping me troubleshoot the issue. \n    Category: customer service\n    ###\n    \n    <<<\n    Inquiry: {inquiry}\n    >>>\n    Category:\n\"\"\"","metadata":{"height":642,"execution":{"iopub.status.busy":"2024-05-04T08:10:19.003001Z","iopub.execute_input":"2024-05-04T08:10:19.003418Z","iopub.status.idle":"2024-05-04T08:10:19.011113Z","shell.execute_reply.started":"2024-05-04T08:10:19.003375Z","shell.execute_reply":"2024-05-04T08:10:19.009923Z"},"trusted":true},"execution_count":11,"outputs":[]},{"cell_type":"markdown","source":"#### Ask Mistral to check the spelling and grammar of your prompt","metadata":{}},{"cell_type":"code","source":"prompt = f\"Please correct the spelling and grammar of \\\nthis prompt and return a text that is the same prompt,\\\nwith the spelling and grammar fixed: {prompt}\"\n\nchat_response = mistral.chat(\n    model=model,\n    messages=[ChatMessage(role=\"user\", content=prompt)]\n)\n\nprint(chat_response.choices[0].message.content)","metadata":{"height":64,"execution":{"iopub.status.busy":"2024-05-04T08:10:55.721356Z","iopub.execute_input":"2024-05-04T08:10:55.721793Z","iopub.status.idle":"2024-05-04T08:11:01.527765Z","shell.execute_reply.started":"2024-05-04T08:10:55.721760Z","shell.execute_reply":"2024-05-04T08:11:01.526444Z"},"trusted":true},"execution_count":13,"outputs":[{"name":"stdout","text":"Please correct the spelling and grammar of this prompt and return a text that is the same prompt, with the spelling and grammar fixed:\n\n\"You are a bank customer service bot. Your task is to assess customer intent and categorize customer inquiries after <<>>>> into one of the following predefined categories:\n\n* Card arrival\n* Change pin\n* Exchange rate\n* Country support\n* Cancel transfer\n* Charge dispute\n\nIf the text doesn't fit into any of the above categories, classify it as:\n\n* Customer service\n\nYou will only respond with the predefined category. Do not provide explanations or notes.\n\n###\nHere are some examples:\n\nInquiry: How do I know if I will get my card, or if it is lost? I am concerned about the delivery process and would like to ensure that I will receive my card as expected. Could you please provide information about the tracking process for my card, or confirm if there are any indicators to identify if the card has been lost during delivery?\nCategory: Card arrival\nInquiry: I am planning an international trip to Paris and would like to inquire about the current exchange rates for Euros as well as any associated fees for foreign transactions.\nCategory: Exchange rate\nInquiry: What countries are getting support? I will be traveling and living abroad for an extended period of time, specifically in France and Germany, and would appreciate any information regarding compatibility and functionality in these regions.\nCategory: Country support\nInquiry: Can I get help starting my computer? I am having difficulty starting my computer, and would appreciate your expertise in helping me troubleshoot the issue.\nCategory: Customer service\n###\n\n<<<\nInquiry: {inquiry}\n>>>\nCategory:\n","output_type":"stream"}]},{"cell_type":"markdown","source":"#### Try out the model","metadata":{}},{"cell_type":"code","source":"chat_response = chat_response.choices[0].message.content\nprompt = chat_response.format(\n        inquiry=\"I am inquiring about the availability of your cards in the EU\"\n    )\nchat_response = mistral.chat(\n    model=model,\n    messages=[ChatMessage(role=\"user\", content=prompt)]\n)\n\nprint(chat_response.choices[0].message.content)","metadata":{"height":98,"execution":{"iopub.status.busy":"2024-05-04T08:11:53.798476Z","iopub.execute_input":"2024-05-04T08:11:53.798876Z","iopub.status.idle":"2024-05-04T08:12:01.191335Z","shell.execute_reply.started":"2024-05-04T08:11:53.798844Z","shell.execute_reply":"2024-05-04T08:12:01.189779Z"},"trusted":true},"execution_count":15,"outputs":[{"name":"stdout","text":"You are a bank customer service bot. Your task is to assess customer intent and categorize customer inquiries after <separator> into one of the following predefined categories:\n\n* Card arrival\n* Change pin\n* Exchange rate\n* Country support\n* Cancel transfer\n* Charge dispute\n\nIf the text does not fit into any of the above categories, classify it as:\n\n* Customer service\n\nYou will only respond with the predefined category. Do not provide explanations or notes.\n\nHere are some examples:\n\nInquiry: \"How do I know if I will get my card, or if it is lost? I am concerned about the delivery process and would like to ensure that I will receive my card as expected. Could you please provide information about the tracking process for my card, or confirm if there are any indicators to identify if the card has been lost during delivery?\"\nCategory: Card arrival\n\nInquiry: \"I am planning an international trip to Paris and would like to inquire about the current exchange rates for Euros as well as any associated fees for foreign transactions.\"\nCategory: Exchange rate\n\nInquiry: \"What countries are getting support? I will be traveling and living abroad for an extended period of time, specifically in France and Germany, and would appreciate any information regarding compatibility and functionality in these regions.\"\nCategory: Country support\n\nInquiry: \"Can I get help starting my computer? I am having difficulty starting my computer, and would appreciate your expertise in helping me troubleshoot the issue.\"\nCategory: Customer service\n\n<separator>\nInquiry: \"I am inquiring about the availability of your cards in the EU.\"\nCategory: Country support\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## Information Extraction with JSON Mode","metadata":{}},{"cell_type":"code","source":"medical_notes = \"\"\"\nA 60-year-old male patient, Mr. Johnson, presented with symptoms\nof increased thirst, frequent urination, fatigue, and unexplained\nweight loss. Upon evaluation, he was diagnosed with diabetes,\nconfirmed by elevated blood sugar levels. Mr. Johnson's weight\nis 210 lbs. He has been prescribed Metformin to be taken twice daily\nwith meals. It was noted during the consultation that the patient is\na current smoker. \n\"\"\"","metadata":{"height":166,"execution":{"iopub.status.busy":"2024-05-04T08:12:06.012800Z","iopub.execute_input":"2024-05-04T08:12:06.013301Z","iopub.status.idle":"2024-05-04T08:12:06.020138Z","shell.execute_reply.started":"2024-05-04T08:12:06.013265Z","shell.execute_reply":"2024-05-04T08:12:06.018663Z"},"trusted":true},"execution_count":16,"outputs":[]},{"cell_type":"code","source":"prompt = f\"\"\"\nExtract information from the following medical notes:\n{medical_notes}\n\nReturn json format with the following JSON schema: \n\n{{\n        \"age\": {{\n            \"type\": \"integer\"\n        }},\n        \"gender\": {{\n            \"type\": \"string\",\n            \"enum\": [\"male\", \"female\", \"other\"]\n        }},\n        \"diagnosis\": {{\n            \"type\": \"string\",\n            \"enum\": [\"migraine\", \"diabetes\", \"arthritis\", \"acne\"]\n        }},\n        \"weight\": {{\n            \"type\": \"integer\"\n        }},\n        \"smoking\": {{\n            \"type\": \"string\",\n            \"enum\": [\"yes\", \"no\"]\n        }}\n}}\n\"\"\"","metadata":{"height":472,"execution":{"iopub.status.busy":"2024-05-04T08:13:25.037152Z","iopub.execute_input":"2024-05-04T08:13:25.037585Z","iopub.status.idle":"2024-05-04T08:13:25.044038Z","shell.execute_reply.started":"2024-05-04T08:13:25.037546Z","shell.execute_reply":"2024-05-04T08:13:25.043058Z"},"trusted":true},"execution_count":21,"outputs":[]},{"cell_type":"code","source":"chat_response = mistral.chat(\n    model=model,\n    response_format={\"type\": \"json_object\"},\n    messages=[ChatMessage(role=\"user\", content=prompt)]\n)\n\nprint(chat_response.choices[0].message.content)","metadata":{"height":47,"execution":{"iopub.status.busy":"2024-05-04T08:13:26.634911Z","iopub.execute_input":"2024-05-04T08:13:26.636177Z","iopub.status.idle":"2024-05-04T08:13:27.266368Z","shell.execute_reply.started":"2024-05-04T08:13:26.636125Z","shell.execute_reply":"2024-05-04T08:13:27.264525Z"},"trusted":true},"execution_count":22,"outputs":[{"name":"stdout","text":"{\"age\": 60, \"gender\": \"male\", \"diagnosis\": \"diabetes\", \"weight\": 210, \"smoking\": \"yes\"}\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## Personalization","metadata":{}},{"cell_type":"code","source":"email = \"\"\"\nDear mortgage lender, \n\nWhat's your 30-year fixed-rate APR, how is it compared to the 15-year \nfixed rate?\n\nRegards,\nAnna\n\"\"\"","metadata":{"height":166,"execution":{"iopub.status.busy":"2024-05-04T08:14:06.968808Z","iopub.execute_input":"2024-05-04T08:14:06.969270Z","iopub.status.idle":"2024-05-04T08:14:06.975782Z","shell.execute_reply.started":"2024-05-04T08:14:06.969235Z","shell.execute_reply":"2024-05-04T08:14:06.974661Z"},"trusted":true},"execution_count":31,"outputs":[]},{"cell_type":"code","source":"prompt = f\"\"\"\n\nYou are a mortgage lender customer service bot, and your task is to \ncreate personalized email responses to address customer questions.\nAnswer the customer's inquiry using the provided facts below. Ensure \nthat your response is clear, concise, and directly addresses the \ncustomer's question. Address the customer in a friendly and \nprofessional manner. Sign the email with \"Lender Customer Support.\"   \n      \n# Facts\n30-year fixed-rate: interest rate 6.403%, APR 6.484%\n20-year fixed-rate: interest rate 6.329%, APR 6.429%\n15-year fixed-rate: interest rate 5.705%, APR 5.848%\n10-year fixed-rate: interest rate 5.500%, APR 5.720%\n7-year ARM: interest rate 7.011%, APR 7.660%\n5-year ARM: interest rate 6.880%, APR 7.754%\n3-year ARM: interest rate 6.125%, APR 7.204%\n30-year fixed-rate FHA: interest rate 5.527%, APR 6.316%\n30-year fixed-rate VA: interest rate 5.684%, APR 6.062%\n\n# Email\n{email}\n\"\"\"","metadata":{"height":404,"execution":{"iopub.status.busy":"2024-05-04T08:14:08.157250Z","iopub.execute_input":"2024-05-04T08:14:08.157769Z","iopub.status.idle":"2024-05-04T08:14:08.165270Z","shell.execute_reply.started":"2024-05-04T08:14:08.157720Z","shell.execute_reply":"2024-05-04T08:14:08.163716Z"},"trusted":true},"execution_count":32,"outputs":[]},{"cell_type":"code","source":"chat_response = mistral.chat(\n    model=model,\n    messages=[ChatMessage(role=\"user\", content=prompt)]\n)\n\nprint(chat_response.choices[0].message.content)","metadata":{"height":47,"scrolled":true,"execution":{"iopub.status.busy":"2024-05-04T08:14:09.157010Z","iopub.execute_input":"2024-05-04T08:14:09.157443Z","iopub.status.idle":"2024-05-04T08:14:11.972557Z","shell.execute_reply.started":"2024-05-04T08:14:09.157409Z","shell.execute_reply":"2024-05-04T08:14:11.970465Z"},"trusted":true},"execution_count":33,"outputs":[{"name":"stdout","text":"Dear Anna,\n\nThank you for reaching out to us regarding our mortgage rates. I'm happy to help you with your inquiry.\n\nFor a 30-year fixed-rate mortgage, our current interest rate is 6.403% with an Annual Percentage Rate (APR) of 6.484%. On the other hand, our 15-year fixed-rate mortgage has an interest rate of 5.705% and an APR of 5.848%.\n\nComparatively, the 15-year fixed-rate mortgage has a lower interest rate and APR than the 30-year fixed-rate mortgage. However, the monthly payments for a 15-year fixed-rate mortgage are typically higher due to the shorter loan term.\n\nI hope this information is helpful to you. If you have any further questions, please don't hesitate to ask.\n\nBest regards,\n\nLender Customer Support\n","output_type":"stream"}]},{"cell_type":"markdown","source":"## Summarization\n\n- We'll use this [article](https://www.deeplearning.ai/the-batch/mistral-enhances-ai-landscape-in-europe-with-microsoft-partnership-and-new-language-models) from The Batch","metadata":{}},{"cell_type":"code","source":"newsletter = \"\"\"\nEuropean AI champion Mistral AI unveiled new large language models and formed an alliance with Microsoft. \n\nWhat’s new: Mistral AI introduced two closed models, Mistral Large and Mistral Small (joining Mistral Medium, which debuted quietly late last year). Microsoft invested $16.3 million in the French startup, and it agreed to distribute Mistral Large on its Azure platform and let Mistral AI use Azure computing infrastructure. Mistral AI makes the new models available to try for free here and to use on its La Plateforme and via custom deployments.\n\nModel specs: The new models’ parameter counts, architectures, and training methods are undisclosed. Like the earlier, open source Mistral 7B and Mixtral 8x7B, they can process 32,000 tokens of input context. \n\nMistral Large achieved 81.2 percent on the MMLU benchmark, outperforming Anthropic’s Claude 2, Google’s Gemini Pro, and Meta’s Llama 2 70B, though falling short of GPT-4. Mistral Small, which is optimized for latency and cost, achieved 72.2 percent on MMLU.\nBoth models are fluent in French, German, Spanish, and Italian. They’re trained for function calling and JSON-format output.\nMicrosoft’s investment in Mistral AI is significant but tiny compared to its $13 billion stake in OpenAI and Google and Amazon’s investments in Anthropic, which amount to $2 billion and $4 billion respectively.\nMistral AI and Microsoft will collaborate to train bespoke models for customers including European governments.\nBehind the news: Mistral AI was founded in early 2023 by engineers from Google and Meta. The French government has touted the company as a home-grown competitor to U.S.-based leaders like OpenAI. France’s representatives in the European Commission argued on Mistral’s behalf to loosen the European Union’s AI Act oversight on powerful AI models. \n\nYes, but: Mistral AI’s partnership with Microsoft has divided European lawmakers and regulators. The European Commission, which already was investigating Microsoft’s agreement with OpenAI for potential breaches of antitrust law, plans to investigate the new partnership as well. Members of President Emmanuel Macron’s Renaissance party criticized the deal’s potential to give a U.S. company access to European users’ data. However, other French lawmakers support the relationship.\n\nWhy it matters: The partnership between Mistral AI and Microsoft gives the startup crucial processing power for training large models and greater access to potential customers around the world. It gives the tech giant greater access to the European market. And it gives Azure customers access to a high-performance model that’s tailored to Europe’s unique regulatory environment.\n\nWe’re thinking: Mistral AI has made impressive progress in a short time, especially relative to the resources at its disposal as a startup. Its partnership with a leading hyperscaler is a sign of the tremendous processing and distribution power that remains concentrated in the large, U.S.-headquartered cloud companies.\n\"\"\"","metadata":{"height":336,"execution":{"iopub.status.busy":"2024-05-04T08:14:14.777940Z","iopub.execute_input":"2024-05-04T08:14:14.778839Z","iopub.status.idle":"2024-05-04T08:14:14.789119Z","shell.execute_reply.started":"2024-05-04T08:14:14.778793Z","shell.execute_reply":"2024-05-04T08:14:14.787553Z"},"trusted":true},"execution_count":34,"outputs":[]},{"cell_type":"code","source":"prompt = f\"\"\"\nYou are a commentator. Your task is to write a report on a newsletter. \nWhen presented with the newsletter, come up with interesting questions to ask,\nand answer each question. \nAfterward, combine all the information and write a report in the markdown\nformat. \n\n# Newsletter: \n{newsletter}\n\n# Instructions: \n## Summarize:\nIn clear and concise language, summarize the key points and themes \npresented in the newsletter.\n\n## Interesting Questions: \nGenerate three distinct and thought-provoking questions that can be \nasked about the content of the newsletter. For each question:\n- After \"Q: \", describe the problem \n- After \"A: \", provide a detailed explanation of the problem addressed \nin the question.\n- Enclose the ultimate answer in <>.\n\n## Write a analysis report\nUsing the summary and the answers to the interesting questions, \ncreate a comprehensive report in Markdown format. \n\"\"\"","metadata":{"height":472,"execution":{"iopub.status.busy":"2024-05-04T08:14:15.911398Z","iopub.execute_input":"2024-05-04T08:14:15.911802Z","iopub.status.idle":"2024-05-04T08:14:15.917480Z","shell.execute_reply.started":"2024-05-04T08:14:15.911768Z","shell.execute_reply":"2024-05-04T08:14:15.916280Z"},"trusted":true},"execution_count":35,"outputs":[]},{"cell_type":"code","source":"chat_response = mistral.chat(\n    model=model,\n    messages=[ChatMessage(role=\"user\", content=prompt)]\n)\n\nprint(chat_response.choices[0].message.content)","metadata":{"height":47,"scrolled":true,"execution":{"iopub.status.busy":"2024-05-04T08:14:23.025168Z","iopub.execute_input":"2024-05-04T08:14:23.025680Z","iopub.status.idle":"2024-05-04T08:14:34.654088Z","shell.execute_reply.started":"2024-05-04T08:14:23.025638Z","shell.execute_reply":"2024-05-04T08:14:34.652972Z"},"trusted":true},"execution_count":37,"outputs":[{"name":"stdout","text":"# Summary:\n\nEuropean AI champion Mistral AI has unveiled new large language models and formed an alliance with Microsoft. The French startup introduced two new closed models, Mistral Large and Mistral Small, and received a $16.3 million investment from Microsoft. Microsoft will distribute Mistral Large on its Azure platform and collaborate with Mistral AI to train bespoke models for customers, including European governments. Mistral AI's new models outperform some competitors on the MMLU benchmark but fall short of GPT-4. The partnership has divided European lawmakers and regulators, with some expressing concerns about data privacy and antitrust issues.\n\n# Interesting Questions:\n\nQ: How does Mistral AI's performance on the MMLU benchmark compare to other prominent models in the industry?\n\nA: Mistral Large achieved an impressive score of 81.2% on the MMLU benchmark, outperforming Anthropic’s Claude 2, Google’s Gemini Pro, and Meta’s Llama 2 70B. However, it still falls short of GPT-4's performance. This demonstrates that Mistral AI's models are competitive in the market and can provide valuable AI capabilities, but there is still room for improvement to catch up with industry leaders.\n\n<81.2% is a strong score, indicating that Mistral Large is a capable model. However, GPT-4 outperforms it, suggesting that Mistral AI still has work to do to reach the top tier of language models.>\n\nQ: What are the potential implications of the partnership between Mistral AI and Microsoft for European data privacy and antitrust regulations?\n\nA: The partnership between Mistral AI and Microsoft has raised concerns among some European lawmakers and regulators about data privacy and antitrust issues. The European Commission plans to investigate the agreement for potential breaches of antitrust law, and members of President Emmanuel Macron’s Renaissance party have criticized the deal's potential to give a U.S. company access to European users' data. These concerns highlight the challenges of balancing the benefits of AI innovation with the need to protect users' privacy and maintain fair competition in the European market.\n\n<The partnership between Mistral AI and Microsoft may face significant regulatory hurdles in Europe, as data privacy and antitrust concerns take center stage. This situation underscores the importance of addressing these issues proactively to ensure the sustainable growth of the AI industry.>\n\nQ: How does Mistral AI's rapid progress and partnership with a leading hyperscaler reflect the current state of the AI market, particularly in terms of processing power and distribution?\n\nA: Mistral AI has made impressive progress in a short time, especially considering its resources as a startup. Its partnership with a leading hyperscaler like Microsoft highlights the tremendous processing and distribution power that remains concentrated in large, U.S.-headquartered cloud companies. This situation suggests that AI startups may need to form strategic partnerships with these dominant players to access the necessary resources and reach a global customer base.\n\n<Mistral AI's experience illustrates the significant advantages that come with partnering with a leading hyperscaler in the AI market. However, this also raises questions about the distribution of power and resources in the industry, and whether adequate opportunities exist for smaller players and new entrants.>\n\n# Analysis Report:\n\nEuropean AI champion Mistral AI has unveiled new large language models and formed an alliance with Microsoft, marking a significant development in the European AI landscape. Mistral AI introduced two new closed models, Mistral Large and Mistral Small, and received a $16.3 million investment from Microsoft. This partnership enables Microsoft to distribute Mistral Large on its Azure platform and collaborate with Mistral AI to train bespoke models for customers, including European governments.\n\nMistral AI's new models have demonstrated strong performance on the MMLU benchmark, outperforming some competitors but still falling short of GPT-4. This achievement highlights the company's rapid progress and competitive position in the market. However, the partnership with Microsoft has raised concerns among some European lawmakers and regulators regarding data privacy and antitrust issues. The European Commission plans to investigate the agreement for potential breaches of antitrust law, and critics have expressed concerns about the potential for a U.S. company to access European users' data.\n\nThese challenges notwithstanding, Mistral AI's experience showcases the significant advantages of partnering with a leading hyperscaler in the AI market. Access to Microsoft's processing power and distribution network provides crucial support for training large models and reaching potential customers worldwide. For Azure customers, this partnership offers a high-performance model tailored to Europe's unique regulatory environment.\n\nIn conclusion, Mistral AI's rapid progress and strategic partnership with Microsoft demonstrate the tremendous processing and distribution power concentrated in large, U.S.-headquartered cloud companies. While this situation offers opportunities for AI startups to scale and innovate, it also raises questions about the distribution of power and resources in the industry, and whether adequate opportunities exist for smaller players and new entrants. Addressing these concerns is crucial for ensuring the sustainable growth and competitiveness of the global AI ecosystem.\n","output_type":"stream"}]},{"cell_type":"markdown","source":"#### Try it out for yourself\n- Feel free to copy-paste text from another article in The Batch, or any other blog or news article.","metadata":{}},{"cell_type":"markdown","source":"## The Mistral Python client\n- Below is the helper function that you imported from helper.py and used earlier in this notebook.\n- For more details, check out the [Mistral AI API documentation](https://docs.mistral.ai/api/)\n- To get your own Mistral AI API key to use on your own, outside of this classroom, you can create an account and go to the [console](https://console.mistral.ai/) to subscribe and create an API key.","metadata":{}},{"cell_type":"code","source":"from mistralai.client import MistralClient\nfrom mistralai.models.chat_completion import ChatMessage   \n\ndef mistral(user_message, \n            model=\"mistral-small-latest\",\n            is_json=False):\n    client = MistralClient(api_key=os.getenv(\"MISTRAL_API_KEY\"))\n    messages = [ChatMessage(role=\"user\", content=user_message)]\n\n    if is_json:\n        chat_response = client.chat(\n            model=model, \n            messages=messages,\n            response_format={\"type\": \"json_object\"})\n    else:\n        chat_response = client.chat(\n            model=model, \n            messages=messages)\n        \n    return chat_response.choices[0].message.content","metadata":{"height":353,"execution":{"iopub.status.busy":"2024-05-04T08:14:44.205112Z","iopub.execute_input":"2024-05-04T08:14:44.205550Z","iopub.status.idle":"2024-05-04T08:14:44.215878Z","shell.execute_reply.started":"2024-05-04T08:14:44.205513Z","shell.execute_reply":"2024-05-04T08:14:44.214286Z"},"trusted":true},"execution_count":38,"outputs":[]},{"cell_type":"markdown","source":"","metadata":{}}]}