fix: improve conclusion

recipes/finetuning_openai_summaries.ipynb

@@ -1161,6 +1161,14 @@
     "Unfortunately, apart from the examples 3 and 9 which are clearly not functionning, it is not obvious which version is better."
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Conclusion\n",
+    "This notebooks shows how to create a LLM fine-tuning project with Kili and Chat GPT 3.5, on a summarization use case. Given the excessive simplicity of the performance metric and the small size of the dataset, the results are not convincing. However, this notebook can be used as a starting point to build a more realistic pipeline, with a more reliable metric and a larger dataset."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,