去年和中信证券完成元宇宙的相关实习，作者开始非常关注人工智能的赛道, 并且猛学了几个月的Python。从VR, NFT等概念到今年的AIGC爆发，人工智能已经成为了我们生活中必不可少的一部分。
Since completing an internship related to Metaverse with CITIC Securities last year, the author has been paying close attention to the Artificial Intelligence (AI) industry and has been furiously studying Python for several months. From concepts such as VR, NFT to the explosion of AIGC this year, AI has become an indispensable part of our lives.
虽然，AI在短时间内还不会达到完全取代人工的地步，但是学会使用AI已经成为了一项必备的技能。通过几个小时的学习曲线就可以提升工作效率。现在应该是能将AI快速使用在工作里的会取代那些依赖传统方式的打工人，不会使用AI的会输给世界的发展。毕竟努力不一定代表有结果~
Although AI won’t be able to completely replace human labor in the short-term, it has become essential to learn how to use AI. With only a few hours of learning curve, one can greatly improve their work efficiency. Now, those who are able to use AI will replace those who rely on traditional working methods, and those who don’t use AI will be left behind in the development of the world. We all know that effort doesn‘t necessarily guarantee a result.
作者非常推荐AI导航软件Futurepedia：https://www.futurepedia.io 已经收录了一千多个AI工具。这些工具可使用的范围覆盖了各行各业。
The author highly recommends the AI navigation software Futurepedia: https://www.futurepedia.io, which has already compiled more than a thousand AI tools. These tools are applicable across a wide range of industries.
除了ChatGPT,下面介绍几个领域常用的AI软件：
文字生成图片：DALL·E、Stable Diffusion、Midjourney、craiyon、Jasper、Imagen、Wombo
文字生成视频：runway、Fliki、synthesia、MetaAI、Google AI、Phenaki
文字生成音频：Play.ht、MURF. AI、RESENBLE.AI、WELLSAID
文字生成文字：Simplifield、Jasper、grammarly、copi.ai
文字生成动画：TREE.Ind.、MDM：Human Motion Diffusion Model
文字生成代码：replitGenerate code、GitHub Copilot
文字生成NFT：LensAI
文字生成3D动画： DreamFusion、CLIP-MESH、GET3D
音频生成文字：descript、AssemblyAI、Whisper
音频生成音频：AudioLM、VOICEMOD
脑图生成文字：speech from brain、non-invasive brain recordings
图片生成文字：neural.love、GPT-3×Image Captions
Apart from ChatGPT, here are a few commonly used AI software:
Text-to-Image：DALL·E、Stable Diffusion、Midjourney、craiyon、Jasper、Imagen、Wombo
Text-to-Video：runway、Fliki、synthesia、MetaAI、Google AI、Phenaki
Text-to-Audio：Play.ht、MURF. AI、RESENBLE.AI、WELLSAID
Text-to-Text：Simplifield、Jasper、grammarly、copi.ai
Text-to-Motion：TREE.Ind.、MDM：Human Motion Diffusion Model
Text-to-code：replitGenerate code、GitHub Copilot
Text-to-NFT：LensAI
Text-to-3D： DreamFusion、CLIP-MESH、GET3D
Audio-to-Text：descript、AssemblyAI、Whisper
Audio-to-Audio：AudioLM、VOICEMOD
Brain-to-Text：speech from brain、non-invasive brain recordings
Image-to-Text：neural.love、GPT-3×Image Captions
本文彩蛋：
作者现在邀请第一位虚拟员工为大家介绍Aries Investment:
Without further ado, the author now invites the first virtual employee to introduce Aries Investment to everyone.
最近已经有很多人讨论过AI的优点，作者要在这里讨论一下AI的缺点:
Let‘s discuss the drawbacks of AI here, as there have already been many discussions about the advantages of AI recently:
- 主要缺点是AI受预训练模型的限制，无法生成完全原创的内容。
- AI容易产生过度普通或重复的回应，并且无法提供特定上下文的信息。
- 要为特定应用程序定制模型可能需要花费大量时间和成本。
- 该模型并不总是准确的，因此可能会产生错误或不当的反应。
- AI在生成新的内容上有很强的优势，比如撰写观点，论文，宣传语等。但是很难替我们描述我们的经历和过去发生故事。毕竟什么也替代不了我们的记忆。
- AI的软件和更新速度非常快。这意味着今日所学的AI技能很快会被取代。所以，在这个AI爆发的时代，我们需要很快速的学习能力，但从一定程度上讲是非常劳累和消耗时间的。有时候表示学不动了~
- 使用AI反而对于使用者的提问能力和操作能力有一定要求。比如，AI作画当给出的描述很粗略时，生成的图片可用度极低。只有非常详尽的描述，才可以生成贴合我们需要的图片。这对于AI使用者的技术有很高要求。需要AI使用者对于语言表述，画作风格等拥有一定的知识储备。ChatGPT的内容生成质量和我们给的问题，资料，信息密切相关。像ChatGPT输出的结果有时很有逻辑，但是一本正经的胡说八道。往往使用者需要像老师一样为学生批改作业，对于使用者的细致和逻辑思维都很有挑战。
- The main drawback is that it is limited by its pre-trained model and cannot generate completely original content.
- It is prone to generating overly generic or repetitive responses, and it cannot provide context-specific information.
- It can be time-consuming and costly to customize the model for specific applications.
- The model is not always accurate, so it may generate incorrect or inappropriate responses.
- AI has many advantages in generating new content, such as expressing opinions, writing papers and slogan–making, but it is difficult to replace our memories to describe our experiences and past stories.
- The pace of AI software and updates is very fast, which means the AI skills learned today will be replaced quickly. Therefore, in this AI-explosive era, we need to learn quickly and it is very tiring and time-consuming.
- Using AI requires certain requirements of the user’s questioning ability and operational ability. For example, when the description given for AI painting is very rough, the generated picture has extremely low usability. Only very detailed descriptions can generate the picture that meets our needs. This requires AI users to have certain knowledge reserves in terms of language expression and painting style. The output of ChatGPT is closely related to the questions, data and information we give. Sometimes the results of ChatGPT are logical, but they are often nonsense. Users often need to be like teachers and correct assignments for students, which challenges the user’s meticulousness and logic thinking.
