智能体的L0～L5分级（Levels of agent automation，智能体自动化级别）
<p class="article-paragraph" data-doubao-translate-traverse-mark="1"><span class="inf-bold">Table 4: Levels of agent automation</span></p><p class="article-paragraph" data-doubao-translate-traverse-mark="1"><span class="inf-bold"><span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">表4：智能体自动化级别</span></span></span></p>
<table class="table table-bordered af-table" cellspacing="0" cellpadding="0" data-mediaid="8516c1d4fdaf4c34bd05245ebbca7b40" data-doubao-translate-traverse-mark="1">
<tbody data-doubao-translate-traverse-mark="1">
<tr data-doubao-translate-traverse-mark="1">
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span class="inf-bold">Level<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">等级</span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span class="inf-bold">Description<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">描述</span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span class="inf-bold">Key characteristics<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">关键特征</span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span class="inf-bold">Examples<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">示例</span></span></span></p>
</td>
</tr>
<tr data-doubao-translate-traverse-mark="1">
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Level 0: Manual (no automation)<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">0级：手动（无自动化）</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">No AI involvement; humans perform all actions<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">无人工智能参与；所有操作均由人类执行</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- No autonomy<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 无自主性</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Tools are passive<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 工具是被动的</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- No perception, decision, or action loop<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 无感知、决策或行动循环</span></span></span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Spreadsheets, command-line tools<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">电子表格、命令行工具</span></span></p>
</td>
</tr>
<tr data-doubao-translate-traverse-mark="1">
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Level 1: Rule-based agents<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">一级：基于规则的智能体</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Agents follow static logic (e.g., if-this-then-that) or rules; no context/environment perception or learning<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">智能体遵循静态逻辑（例如，若此则彼）或规则；无上下文/环境感知或学习能力。</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Deterministic logic<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 确定性逻辑</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- No adaptation<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 无适应性</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- No memory use <span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__">- 不使用记忆</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- No goal-awareness<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 无目标意识</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Human-in-the-loop for decision-making<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 决策中的人在回路</span></span></span></span></span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Traditional RPA tools, a rules-based chatbot for simple Q&A, an interactive voice response (IVR) system that routes calls based on keywords<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">传统的机器人流程自动化（RPA）工具、用于简单问答的基于规则的聊天机器人、一种基于关键词来转接电话的交互式语音应答（IVR）系统</span></span></p>
</td>
</tr>
<tr data-doubao-translate-traverse-mark="1">
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Level 2: Reactive agents <span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__">二级：反应式智能体</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Agents can adapt responses based on defined conditions/context; use a narrow set of predefined tools (e.g., APIs, databases) to accomplish tasks but within well-defined workflows initiated by humans<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">智能体可以根据定义的条件/上下文调整回复；使用少量预定义的工具（如应用程序编程接口、数据库）来完成任务，但这些任务需在由人类发起的明确工作流程内进行。</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Can process structured/unstructured input<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 能够处理结构化/非结构化输入</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Uses ML for input understanding; e.g., natural language processing (NLP)<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 使用机器学习来理解输入；例如，自然语言处理（NLP）</span></span></span></span></p>
<p data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Executes decisions based on context and rigid workflows<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 根据上下文和严格的工作流程执行决策</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Can execute decisions based on certain conditions (e.g., location)<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 可以根据特定条件（例如位置）执行决策</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- No long-term planning<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 无长期规划</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Narrow domain-specific task autonomy<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 特定领域的窄任务自主性</span></span></span></span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Intelligent RPA tools, chatbots that can answer common questions using a knowledge base, a location-based assistant (e.g., a local restaurant finder app)<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">智能RPA工具、能够利用知识库回答常见问题的聊天机器人、基于位置的智能助手（例如，一款本地餐厅查找应用程序）</span></span></p>
</td>
</tr>
<tr data-doubao-translate-traverse-mark="1">
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Level 3: Conditional autonomous agents<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">3级：有条件的自主智能体</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">AI can pursue goals with intermediate planning under constraints<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">人工智能可以在约束条件下通过中间规划来实现目标</span></span></p>
<p data-doubao-translate-traverse-mark="1">May require human intervention for novel or complex situations<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">对于新出现或复杂的情况，可能需要人工干预。</span></span> </p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Task decomposition<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 任务分解</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Reasoning and feedback loops<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 推理与反馈循环</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Domain-bounded<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 领域受限的</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Uses symbolic or learned policy for decisions<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 使用符号化策略或习得策略进行决策</span></span></span></span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">GenAI agents, and copilots<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">生成式人工智能智能体与副驾驶</span></span></p>
</td>
</tr>
<tr data-doubao-translate-traverse-mark="1">
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Level 4: Strategic autonomous agents<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">第4级：战略型自主智能体</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Complex reasoning and task decomposition to proactively achieve complex goals with minimal or no human oversight<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">复杂推理和任务分解，以便在最少或无需人工监督的情况下主动实现复杂目标</span></span></p>
<p data-doubao-translate-traverse-mark="1">Highly adaptable and can function in dynamic environments<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">适应性强，能够在动态环境中运行</span></span></p>
<p data-doubao-translate-traverse-mark="1"> </p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Use advanced model reasoning and capabilities (e.g., multi-modal)<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 使用先进的模型推理和能力（例如，多模态）</span></span><span data-doubao-translate-traverse-mark="1"> <br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">-Advanced tool use <span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__">- 高级工具使用</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Proactive goal setting<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 主动设定目标</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Long-horizon planning<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 长期规划</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Can operate with partial observability and uncertainty<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">- 可以在部分可观测和不确定的情况下运行</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Still domain bound<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 仍然受领域限制</span></span></span></span></span></span></span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">A domain-specific AI agent that can proactively plan and execute a workflow end-to-end (e.g., Salesforce Agentforce and Devin, a software engineering agent from Cognition AI)<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">一种特定领域的人工智能智能体，它可以主动地端到端规划并执行工作流程（例如，Salesforce Agentforce以及来自Cognition AI的软件工程智能体Devin）</span></span></p>
<p data-doubao-translate-traverse-mark="1">Multi-agent systems (e.g., AutoGPT)<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">多智能体系统（例如，AutoGPT）</span></span></p>
</td>
</tr>
<tr data-doubao-translate-traverse-mark="1">
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Level 5: General autonomous agents—artificial general intelligence (AGI) agents<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">第5级：通用自主智能体——通用人工智能（AGI）智能体</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Agents can operate across multiple domains, autonomously set their own goals, and use adaptive strategies<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">智能体可以跨多个领域运作，自主设定自己的目标，并使用适应性策略。</span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Multi-domain<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 多领域</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Meta-reasoning<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 元推理</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Adaptive self-learning<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 自适应自主学习</span></span><br data-doubao-translate-traverse-mark="1"><span data-doubao-translate-traverse-mark="1">- Critical value alignment<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">- 关键价值对齐</span></span></span></span></span></span></p>
</td>
<td class="inf-top" data-doubao-translate-traverse-mark="1">
<p data-doubao-translate-traverse-mark="1">Hypothetical AGI agents, experimental<span class="__Cici__translate__"><span class="__Cici_translate_translated_inject_node__ __Cici_translate_need_wrapper__">假设的通用人工智能智能体，实验性的</span></span></p>
</td>
</tr>
</tbody>
</table>
<p class="article-paragraph" data-doubao-translate-traverse-mark="1">Source: Omdia<span class="__Cici__translate__"> <span class="__Cici_translate_translated_inject_node__">来源：Omdia</span></span></p>
</div>
<div id="MySignature" role="contentinfo">
<br><br>
<div>
<p style="text-align:center;">
<img src="https://s3.ax1x.com/2020/12/02/D5XzUP.png" height=80px>
<br>
<p style="text-align:center;">如果这篇文章帮助到了你，你可以请作者喝一杯咖啡</p>
<p style="text-align:center;"></p>
<img src="https://img2020.cnblogs.com/blog/938105/202010/938105-20201013171833350-276632012.png" height=250px align="center"/>
</div><br><br>
来源：https://www.cnblogs.com/sddai/p/18973208
頁:
[1]