The first thing I did this morning was argue with an AI (\u274c) \u2014 I mean, have a rational discussion (\u2705) \u2014 about the differences between AI agents and traditional apps calling AI:<\/p>\n\n\n\n
(p.s. The following list items are from ChatGPT, but all the examples and explanations were written by a real human. Enjoy! \\:D)<\/p>\n\n\n\n
<\/p>\n\n\n\n
(Fixed workflow vs. Flexible and extensible workflow)<\/strong><\/p>\n\n\n\n Traditional apps calling AI can certainly use AI to accomplish specific goals (e.g., retrieving today\u2019s news), but the process is fixed. If you want to add a new feature (e.g., get today\u2019s news and filter out news related to Trump), you have to write new code to support it.<\/p>\n\n\n\n On the other hand, an AI agent, as long as the new task \u2014 like “filtering Trump-related news” \u2014 can still be completed using AI or existing supported tools (Internet, search, etc.), does not require manually adding new code to change the current workflow.<\/p>\n\n\n\n (Stateless vs. Stateful)<\/strong><\/p>\n\n\n\n Traditional apps calling AI are basically stateless \u2014 each execution is independent and not affected by previous runs. AI agents, however, are fundamentally designed with memory (which may be short-term or long-term).<\/p>\n\n\n\n Note: While developers can add memory manually when using traditional apps, AI agents have this memory mechanism built into their infrastructure \u2014 no need for developers to design a separate \u201cmemory system.\u201d<\/p>\n\n\n\n (Fixed output vs. Dynamic and differentiated output)<\/strong><\/p>\n\n\n\n Traditional apps can only display results designed by developers (e.g., listing all Trump-related news headlines), but AI agents can dynamically generate different types of output based on the input (e.g., one time it shows news headlines; another time, it extracts all the dates mentioned in the news).<\/p>\n\n\n\n (Fixed task steps vs. Flexible and adaptive steps)<\/strong><\/p>\n\n\n\n Traditional apps follow fixed task procedures (e.g., retrieve all news \u2192 filter Trump-related news \u2192 extract the headlines). In contrast, AI agents can alter their workflow based on input or environment (e.g., extract all news headlines \u2192 filter for Trump-related ones). Each execution might differ slightly, but the end goal \u2014 getting all Trump-related headlines \u2014 remains the same.<\/p>\n\n\n\n Remarks:<\/strong><\/p>\n\n\n\n AI agents are more extensible than traditional apps or websites because they are built on AI models that can interpret input (e.g., even if you give incorrect instructions, they can often still understand you or fix your intent) and utilize existing tools (e.g., calling APIs or using API schemas) to achieve variable tasks and flexible workflows.<\/p>\n\n\n\n The downside of this flexibility and extensibility is also clear: it\u2019s impossible to guarantee that the output is always correct or as expected.<\/p>\n\n\n\n After all, if the AI model crashes or suddenly becomes “dumb” (i.e., loses its ability to reason properly), you’ll likely end up going back to the traditional app to get the reliable result you originally wanted.<\/p>\n\n\n\n <\/p>\n\n\n\n Below is Chinese version:<\/p>\n\n\n\n \u65e9\u8d77\u7b2c\u4e00\u4ef6\u4e8b\u662f\u8ddfAI\u5435\u67b6(X)\u7406\u6027\u8a0e\u8ad6(O)\u95dc\u65bcAI agent\u548c\u50b3\u7d71 App \u547c\u53eb AI\u7684\u5dee\u5225:<\/p>\n\n\n\n (p.s. \u4ee5\u4e0b\u9664\u4e86\u9805\u76ee\u662fChatGPT\u7d66\u7684\uff0c\u4f8b\u5b50\u8207\u8aaa\u660e\u7686\u70ba\u771f\u4eba\u64b0\u5beb\uff0c\u8acb\u5b89\u5fc3\u670d\u7528:D)<\/strong><\/p>\n\n\n\n (1) \u6d41\u7a0b\u4e0a\u7684\u5dee\u5225:<\/strong><\/p>\n\n\n\n (\u56fa\u5b9a\u6d41\u7a0b vs. \u5f48\u6027\u53ef\u64f4\u5145\u7684\u6d41\u7a0b)<\/p>\n\n\n\n \u50b3\u7d71app\u547c\u53ebAI\uff0c\u96d6\u7136\u4e00\u6a23\u53ef\u4ee5\u5229\u7528AI\u53bb\u9054\u6210\u4efb\u52d9\u76ee\u6a19(ex: \u67e5\u8a62\u4eca\u5929\u65b0\u805e)\uff0c\u4f46\u90fd\u662f\u56fa\u5b9a\u6d41\u7a0b\u3002\u5982\u679c\u60f3\u52a0\u4e0a\u65b0\u529f\u80fd(ex: \u67e5\u8a62\u4eca\u5929\u65b0\u805e\uff0c\u4e26\u627e\u51fa\u8ddf\u5ddd\u666e\u6709\u95dc\u7684\u65b0\u805e)\uff0c\u5c31\u9700\u8981\u5beb\u65b0\u7a0b\u5f0f\u78bc\u9054\u6210\u76ee\u6a19\u3002<\/p>\n\n\n\n \u53cd\u4e4b\uff0cAI agent\u53ef\u4ee5\u5728\u78ba\u4fdd\u300c\u627e\u51fa\u8ddf\u5ddd\u666e\u6709\u95dc\u7684\u65b0\u805e\u300d\u9019\u500b\u65b0\u529f\u80fd\u4e00\u6a23\u53ef\u4ee5\u900f\u904eAI or \u5176\u4ed6\u65e2\u6709\u652f\u63f4\u7684\u5de5\u5177(Internet, search)\u9054\u6210\uff0c\u5c31\u4e0d\u7528\u624b\u52d5\u52a0\u4e0a\u65b0\u7a0b\u5f0f\u78bc\u6539\u8b8a\u65e2\u6709\u6d41\u7a0b\u3002<\/p>\n\n\n\n (2) \u8a18\u61b6\u4e0a\u7684\u5dee\u5225:<\/strong><\/p>\n\n\n\n (stateless vs. stateful)<\/p>\n\n\n\n \u50b3\u7d71app\u547c\u53ebAI\u57fa\u672c\u4e0a\u662fstateless\uff0c\u6bcf\u6b21\u7684\u57f7\u884c\u90fd\u662f\u7368\u7acb\u7684\uff0c\u4e0d\u6703\u53d7\u5230\u524d\u4e00\u6b21\u5f71\u97ff\u3002\u800cAI agent\u7684\u6839\u672c\u8a2d\u8a08\u5c31\u5305\u542b\u8a18\u61b6\u72c0\u614b(\u53ef\u80fd\u662f\u77ed\u671f\u8a18\u61b6\u6216\u9577\u671f\u8a18\u61b6)\u3002<\/p>\n\n\n\n Note: App\u547c\u53ebAI\u5f8c\uff0cApp\u958b\u767c\u8005\u53ef\u4ee5\u81ea\u5df1\u52a0\u4e0a\u8a18\u61b6\u529f\u80fd\uff0c\u4f86\u78ba\u4fddstate\u88ab\u4fdd\u7559\uff0c\u53ea\u662fAI model\u539f\u751f\u7684infrastructure\u5c31\u6709\u652f\u63f4\uff0c\u4e0d\u5fc5\u984d\u5916\u8a2d\u8a08\u300c\u8a18\u61b6\u529f\u80fd\u300d<\/p>\n\n\n\n (3) \u884c\u52d5\u7d50\u679c\u4e0a\u7684\u5dee\u5225:<\/strong><\/p>\n\n\n\n (\u56fa\u5b9a\u7684\u986f\u793a\u7d50\u679c vs. \u5f48\u6027\u5177\u5099\u5dee\u7570\u5316\u7684\u986f\u793a\u7d50\u679c)<\/p>\n\n\n\n \u50b3\u7d71app\u53ea\u80fd\u986f\u793a\u958b\u767c\u8005\u8a2d\u8a08\u7684\u7d50\u679c(ex: \u986f\u793a\u51fa\u6240\u6709\u5ddd\u666e\u65b0\u805e\u7684\u6a19\u984c)\uff0c\u4f46\u662fAI agent\u53ef\u4ee5\u6839\u64da\u4e0d\u540c\u7684input\uff0c\u6bcf\u6b21\u986f\u793a\u5177\u5099\u5dee\u7570\u5316\u7684\u7d50\u679c(ex: \u4e00\u6b21\u662f\u986f\u793a\u65b0\u805e\u7684\u6a19\u984c\u3001\u53e6\u3127\u6b21\u662f\u986f\u793a\u65b0\u805e\u4e2d\u63d0\u5230\u7684\u6240\u6709\u65e5\u671f)<\/p>\n\n\n\n (4) \u4efb\u52d9\u8a2d\u8a08\u65b9\u5f0f\u4e0a\u7684\u5dee\u5225:<\/strong><\/p>\n\n\n\n (\u56fa\u5b9a\u9054\u6210\u4efb\u52d9\u6b65\u9a5f vs. \u5f48\u6027\u5177\u8b8a\u5316\u6027\u7684\u4efb\u52d9\u6b65\u9a5f )<\/p>\n\n\n\n \u50b3\u7d71app\u7684\u6240\u6709\u4efb\u52d9\u90fd\u6709\u56fa\u5b9a\u6d41\u7a0b(ex: \u627e\u51fa\u6240\u6709\u65b0\u805e -> \u627e\u51fa\u6240\u6709\u5ddd\u666e\u7684\u65b0\u805e -> \u627e\u51fa\u9019\u4e9b\u65b0\u805e\u7684\u6a19\u984c)\u3002AI agent\u53ef\u4ee5\u6839\u64dainput\u3001\u74b0\u5883\u4e0d\u540c\uff0c\u7522\u751f\u4e0d\u540c\u7684\u5de5\u4f5c\u6d41\u7a0b(ex: \u627e\u51fa\u6240\u6709\u65b0\u805e\u7684\u6a19\u984c -> \u627e\u51fa\u7576\u4e2d\u6240\u6709\u5ddd\u666e\u65b0\u805e\u7684\u6a19\u984c)\uff0c\u6bcf\u4e00\u6b21\u7684\u57f7\u884c\u65b9\u5f0f\u53ef\u80fd\u7565\u6709\u4e0d\u540c\uff0c\u4f46\u6700\u7d42\u76ee\u7684\u90fd\u662f\u627e\u51fa\u6240\u6709\u5ddd\u666e\u65b0\u805e\u7684\u6a19\u984c\u3002<\/p>\n\n\n\n \u5099\u8a3b:<\/strong><\/p>\n\n\n\n AI\u4e4b\u6240\u4ee5\u6bd4\u65e2\u6709\u7684app\u6216\u7db2\u9801\u7a0b\u5f0f\u5177\u5099\u64f4\u5145\u7684\u80fd\u529b\uff0c\u662f\u56e0\u70ba\u5efa\u7acb\u5728AI model\u5c0d\u65bcinput\u5177\u5099\u89e3\u91cb\u80fd\u529b(p.s. \u4f60\u5c31\u7b97\u4e0b\u932f\u6307\u4ee4\uff0c\u5b83\u9084\u80fd\u5e6b\u4f60\u6821\u6b63\u3002\u52a0\u5165\u65b0\u7684\u8981\u6c42\uff0cAI\u4e00\u6a23\u80fd\u89e3\u91cb\uff0c\u4e0d\u50cf\u76ee\u524d\u524d\u7aef\u7684code\uff0cinvalid input\u5c31\u76f4\u63a5\u639b\u7d66\u4f60\u770b\u6216\u5446\u5446\u7684\u986f\u793a\u7121\u6cd5\u8655\u7406) \u548c\u65e2\u6709\u652f\u63f4\u5de5\u5177\u7684\u80fd\u529b(p.s. \u900f\u904e\u547c\u53eb\u652f\u63f4\u5de5\u5177\u63d0\u4f9b\u7684api or api schema)\u9054\u6210\u8b8a\u5316\u6027\u7684\u4efb\u52d9\u6b65\u9a5f\u548c\u53ef\u64f4\u5145\u6027\u7684\u6d41\u7a0b\u3002<\/p>\n\n\n\n \u9019\u500b\u5f48\u6027\u8207\u64f4\u5145\u6027\u7684\u7f3a\u9ede\u4e5f\u5f88\u660e\u986f\uff0c\u5c31\u662f\u7121\u6cd5\u300c\u4fdd\u8b49\u300d\u6bcf\u6b21\u51fa\u4f86\u7684\u7d50\u679c\u4e00\u5b9a\u662f\u7b26\u5408\u9810\u671f\u6216\u662f\u5c0d\u7684\u3002<\/p>\n\n\n\n \u7562\u7adfAI model\u5982\u679c\u7a81\u7136\u639b\u6a5f\u6216\u7a81\u7136\u8b8a\u300c\u7b28\u300d(p.s. \u5931\u53bb\u4e86\u826f\u597d\u89e3\u91cb\u80fd\u529b)\u7684\u72c0\u6cc1\u4e0b\uff0c\u4f60\u5c31\u53ea\u80fd\u4e56\u4e56\u5730\u8dd1\u56de\u53bb\u7528app\u53bb\u5f97\u5230\u4f60\u9810\u671f\u7684\u7d50\u679c\u3002<\/p>\n","protected":false},"excerpt":{"rendered":" The first thing I did this morning was argue with an AI […]<\/p>\n","protected":false},"author":139426785,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_coblocks_attr":"","_coblocks_dimensions":"","_coblocks_responsive_height":"","_coblocks_accordion_ie_support":"","_postgate_lock_override":"inherit","_postgate_lock_mode":"inherit","_postgate_preview_mode":"inherit","_postgate_preview_limit":0,"_postgate_product_ids":[],"advanced_seo_description":"","jetpack_seo_html_title":"","jetpack_seo_noindex":false,"jetpack_seo_schema_type":"","_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":true,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_wpcom_ai_launchpad_first_post":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"{title}\n\n{excerpt}\n\n{url}","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2},"_wpas_customize_per_network":false,"jetpack_post_was_ever_published":false},"categories":[763194612,55138170,1],"tags":[],"post_folder":[],"class_list":["post-12189","post","type-post","status-publish","format-standard","hentry","category-ai-related","category-technology-sharing","category-uncategorized"],"jetpack_publicize_connections":[],"jetpack_likes_enabled":true,"jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/paR5vw-3aB","jetpack-related-posts":[],"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=\/wp\/v2\/posts\/12189","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=\/wp\/v2\/users\/139426785"}],"replies":[{"embeddable":true,"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=12189"}],"version-history":[{"count":11,"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=\/wp\/v2\/posts\/12189\/revisions"}],"predecessor-version":[{"id":12201,"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=\/wp\/v2\/posts\/12189\/revisions\/12201"}],"wp:attachment":[{"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=12189"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=12189"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=12189"},{"taxonomy":"post_folder","embeddable":true,"href":"https:\/\/1989lifejourney.com\/index.php?rest_route=%2Fwp%2Fv2%2Fpost_folder&post=12189"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}(2) Difference in State:<\/strong><\/h4>\n\n\n\n
(3) Difference in Output Behavior:<\/strong><\/h4>\n\n\n\n
(4) Difference in Task Steps:<\/strong><\/h4>\n\n\n\n
\n\n\n\n