Future & Remix 论坛回放 2 | AI and the Future of Media Consumption
45:02 Share

用 声音。

碰撞 世界。

生动活泼。 在 第一个 panel 里面 我们 聊 了 A I 显著 的 降低 了 内容 创作者 们 的 门槛， 让 更 多人 有 机会 成为 创作者， 从而 生产 出 更多 有趣 的 内容。 所以 我们 第二个 panel 的 主题 就 叫做 A I 与 内容 消费 的 目的。 就是 想要 与 嘉宾 们 一起 探讨 A I 如何 改变 了 我们 生产 和 摄取 内容 的 方式， 以及 我们 要 如何 应对 这个 信息 爆炸 的 时代。

这个 panel 的 主持人 是 容 慧， 她 在 一家 美元 基金 负责 市场 方面 的 工作， 也是 一档 以 A I 为 主题 的 播客 十字路口 的 主播。 我们的 嘉宾 也是 来自于 不同 领域， 像是 尤 兰达， 她 目前 在 香港大学 任教， 之前 有 很多年 的 新闻媒体 的 从业 经验。 因此 对 A I 和 新闻 行业 的 影响 有 很深的 洞察。 另 一位 嘉宾 是 林毅， 他是 在 B 站 拥有 百万 粉丝 的 up 主， 也 曾经 做客 我们的 科技 早知道， 在 wwdc 然后 intelligence 那 期 里面 做 过 我们的 客户 嘉宾。 另外 一位 在 产业界 的 嘉宾， 我们 请来 了 美图 的 C F O gary， 他 来 聊 一 聊 他 从 消费者 那 端 观察 到 的 A I 应用 与 需求。

我 还是 先来 跟 大家 分享 一下， 我在 现场 感受 最深 的 一些 议题 和 他们的 发言。 开场 的 每个 嘉宾 都 提到 了 他们 现在 的 不同 的 领域 以前 不能 做， 但是 A I 让 他们 这些 成为 可能 的 一些 案例。 比如 有 landa 提到 了 路透社 的 一个 叫做 news tracer 的 一个 工具。 这个 工具 可以 实时 的 从 社交 媒体 上 抓取 和 识别 突发 新闻， 显著 提高 了 新闻 的 收集 和 报道 的 速度。 A I 也可以 高效 的 整理 海量 的 政府 文件， 并 让 它们 变得 更加 简单 易读。 Gary 也 提到 了 美图秀秀 利用 A I 也 开发 了 最近 一些 很 受欢迎 的 图片处理 的 功能。

作为 B 站 的 up 主 林毅， 他 观察 到 的 A I 的 广泛应用， 使得 视频 的 创作 变得 更加 民主化 了。 也就是说 现在 每个 人都 可以 借助 A I 来 创作 自己的 作品， 表达 自己的 观点。 从 内容 生产 角度 的 方面 来看， 其实 是 一件 非常 好的事情。 从 消费 这 一端 来看 的话， 就有 了 更多 的 内容。 这样 也 带来 了 一些 挑战， 我们 如何 甄别 和 挑选 高质量 的 内容 就成 了 很大 的 一个 问题 了。

好了， 以上 就是我 对 这个 panel 的 一些 感受 和 观察， 更多 有意思 的 故事 和 细节 就会 在 我们的 节目 中成 现 了。 在 节目 正式 开始 之前， 我 还是 要 感谢 一下 我们的 合作伙伴。 这一次 的 联合 发起方 Z E A B A L A B ultra 康康， 然后 还有 我们的 支持 单位 香港 投资 推广 署， 包括 我们的 赞助 方 图拉斯 以 鹰牌 花旗参。 好了， 下面 就是 我们 今天 的 节目。

So they will introduce themselves respectively. But if you look into their background, uh, this panel is kind of like an industry chain, a conscience crature productivity to provider, observer, researcher and policy researcher. Yeah uh, I if count me as a content crature as well. Okay, so my uh I will uh I will ask my panels to introduce themselves respectively and also can you uh introduced yourself and tell us about the most are not worthy changes A I has brought to your industry and your job, uh, in the past ten years, especially in the past year. Start with your landa.

Hi everyone, this is grandma teaching the of unity of hank. And so the university has been really leading but were not that bad either. Um so um I used to be a journalist practitioner.

Uh I worked in reuters and S M P here in home 卡 recently just joined H K U。 But I had a very long dear in the past decade at work night night on digital policy. That's why um I was introduce and policy research as well. So um in terms of the changes in the industry, i'll be fox more on the new world journalism industry put for the past decade. I will say that this has been playing a very catching up mode.

Um and I think the last year, well, now you put in that way ten years ago, I was still at reuters, the news agency, and at that time we just started looking to um A I and to had a small very small I N D team summary in the north amErica but then at that time, that's like really cool and mysterious something that only a very few people are looking to, right? And now like I think the JoNathan m industry has been pushed into, okay, we have to do something now. Now that everyone become a content creator, everyone become an information provider.

Now what do we do? The the old power dynamics of publishers on the information providing service is breaking down. So I think the industry is really struggling with that. Let me stop the here and pass on the next speaker.

Um thank you your lana my name is gary um and I am the chip financial officer at me too and make you is obviously uh a mobile APP company。 Uh now we have about two hundred sixteen million and monthly active users globally. Uh one third of each is actually outside of england, china.

Um we started pretty much as an APP company that you know try to become a very simplify version of photoshop, focusing on helping users to create you know added face, you making IT more remove permission and all that. We first find that as A P C application. And then in two thousand love and we become a mobile company.

And in two thousand and thirteen, we actually start our own A I lab uh called M T lab. And at that point in time, I think a computer vision is is the most important um A I technology that we have. And we we used to sort of track the points on the face so that you know you can add uh the face very easily um with just one click think uh fast forward um to you know the past one to two years.

I think a generative A I has a really broaden our a company scope um into helping creators um to do a lot of um different applications. Uh for example, right now we have an APP called x design, which is pretty much like canvas, but there's focusing on the e commerce side so users can very easily create a background for A E commerce photo. And then we also have an APP called ki pie， which is are helping your broadcasters or video makers to easily edit their are scripts, and then the video will actually stitch together. You can also add some effect to make IT looks very much like this reality show effects, highlighting the points that you want to say so much. So I think generative A I has brought sort um into helping to create economy a little bit of myself. I feel very compelled to talk about something related to music because of the other people in two thousand and twenty, pretty twenty three, I didn't train a lot of A I, but I train myself playing classical piano and IT was actually very painful process but ended up my able to finish the L T C L programs actually quite have really echo with a lot of the previous pilots view that um you know learning instruments do very important and there is a some some human elements that that really cannot be um replaced by A I at the moment but in the future who knows right let me just pass .

on to uh mr then uh hi guys uh my name is e and i'm a constant creator. So we are running a uh 哔哩哔哩 channel and the youtube channel Michael founders is on the uh sir row here and yeah you can wave to yeah and uh like uh our 哔哩哔哩 china has like one million subscribers but like a million china。 So uh that's basically the population of just one stress. So is no big deal.

So uh our uh our youtube channel has like uh two hundred thousand subscribers and our content is basically focused on uh our uh uh software projects is just like a presents the process of developing its and so we tell our development story, videos and we also uh as our channel growth， we have more opportunities to uh may be talk with uh uh uh uh industrial leaders and a listen to their opinions, maybe make some interviews. So we are also uh doing some of the jobs of uh, journalists. So that's basically what we are doing.

And so uh as for the questions, so um ten years might be too much for me. So because like ten years ago, i'm still undergraduate. So uh if if if we talk about like the the influence of A I on like a in in the scope of ten years on me, it's like a uh IT shapes like my decision of my of my a major selection because like I was still stabbing at that time.

And uh that's one like the the alphago was so popular and I decided to choose uh machine learning as my major uh in uh in my uh graduate study and like uh for the past year. I think uh what i'm observing is that uh the consent uh consent creating process is more and more uh is getting civilized, is is getting like a more and more like, uh, common people are uh using this tool to present their ideas like a previously, if we want to make some films， we want to make some cool things. IT requires like a huge a budget, but not like everybody could have their innovation ideas and presented through A A tools。

IT might not be like that high quality, but uh it's uh they it's it's like a change from zero one. It's like they they begin to have this tool to express themselves. So we are have uh what we uh what I say is that a lot of interesting contents are a pouring into the community like a uh IT IT IT might be uh more chAllenging for us to select, but uh IT also gives more creative ideas the chance to pop out. I so yeah, that's what i'm seen.

Yeah I have a personal story to share that I think the biggest change A I has made to me in my life was starting my podcast, uh, six months ago because I I yeah I told my colleagues about this joke too.

I started past to study AI because I am afraid of losing my as a marketing, as a marketing person yeah so, uh yeah I was a reporter too so I can echo a lot about what you lend a set and am now a podcast producer and like account, uh, as a contents crater as well. So yeah, what 0 is that I I have a lot of feelings too。 So the next part we will talk about, to what extent has A I penetrated the production side of information?

I think, yeah, all of us have a lot have a lot of things to say. So the first question is ah please please uh h please share one to two uh stories uh the innovations that and I has made our lives possible that was previously impossible and for and also what previous on sydney have been fulfilled. Yeah you like .

yeah I can start I I just kind of said the media industry has been playing catch up the game. But just to give credit to some of the chinese in the industry, right? So there has been some agency revolutionary innovations but there has been some um good attempts. Um i'll give two examples. Why is more on the news gathering site? And um so we has had a product I think in seventeen or eighteen which is called news tracer and it's basically doing something that a Normal human being, a reporter will not be able to do basic cripes, seven hundred million daily trees from back then is still a twitter and basically it's identifying what are the breaking news.

And then um so in the news business, the old business model is that the faster the Better right? So basically use A I able to actually scrape all that data and identify the breaking news and give the lead to the reporters for them to identify um things happening like A A what fire kind of happening or um things that if it's just a reporter saying the newsroom may not be able to identify very quickly. So give to us for an agency that win um the completion by time, like something around five minutes to twenty minutes time um competed competitors.

So um that's not necessarily felt by the audience right um because for you ultimately you get the news and you may not really care if it's a five minute difference or try to mean a difference, but IT does matter for the business. And the other example when I give is um from the more production site. And again, that may not be something that the readers actually understand how much has going has been going on the backend.

Um it's actually happening a lot in terms of how A I S used to process thousands, if not millions of documents um for some news organization. So there was one um um U S based organization is a kind of a smaller one, but mostly working on massage journal m and they're able to actually use Jennifer I to summarize the policy documents from all the states across the united states and make them more um easily restful by the readers. So that's another attempt basically that something if you use a um reporter or Anita to do that takes a lot of human time and energy.

Um so one one code I don't remember which editor was that by one of the um biggest news agency in the U S. He said use A I S like one of your best interns, or like a hundreds of interns, right? They're doing a lot of work for you, but you need to train them properly.

So I think these are the two examples. One is on the new year inside and the on. Production site um which from the receiving end you may not really feel that difference um but that's happening um for some newroom and i'll touch on the distribution side maybe later um but I think that's actually one of the changes that in happening most.

So um they said now all the Normal people can start create content. So the old business model of journals m is breaking down, right were not the only competing with um like redos competing ribot, but we're competing with all the thousands, millions, billions people out there. So then how do Normal people like you and me actually receive the information, access the information? How does the information flow works these days? That's actually more a question for everyone to figure out.

Um .

okay. So the next few minutes is going to a sound a little bit like and the further because what we do is the photo and video entering. And obviously A I has a lot of things that changes the way how we added photo. So the first thing, for example, is if there's like a random person you want to remove from the photo, you can use A I just highlight that and then the person will be gone and the background will be generated as if the photo is it's no original. And so there are nit sort of functionalities that most people will use IT.

But if we really need IT as IT would be awesome, for example, like my teeth that say it's not very um in english, not very it's not very aligned and initially there's no way for you to fix that and thus you wear braces right but now our APP can actually you know your teeth and then make IT looks very nice. Your airline, you know you can really do much about IT unless you go to to see A A surgeon but again, our APP can do IT um and for video as well like you know we would have thought that initially no video could be a more you know representative of how you look at you look like but now you can add a lot of you know cosmetics. You can change how you look and then you can you just need to edit one photo and then the effect we applied to the entire video.

And no matter how you move, that edit won't go away. You would just stick to your face and then you look perfect for 有时候 a lot of these things were made possible by A I。 But I wanted emphasize is not just generative A I um no, traditional A I is also very important if you look at out a percentage of users and that uses one or more of the ai, aliis is over eighty seven percent. So our m even though we don't people don't really see IT as an A I APP, but a lot of users are actually using A I functionalities .

uh just like uh ulana said, uh uh like uh uh we are we are having this uh this opportunity of having like more a quick processing of like uh of uh of content and like I I I hate my my interns。 Like, but like A A I could be like a response like more quickly. But like, I hate A I too, because I is so hard to talk with them.

I love, I love the technology, but like, I hate like the way to talk to them and like, uh but but that is a is a new opportunity. Like we have this power of having like so much uh so much like tools to have at our use and we can have quick response from them. I like grisi. We are having this like new uh new technology like image processing and video processing. And uh my point is from like what I said is we are seeing more and more a people are having the capability to present their idea to the world with this new A I to and from uh but from my a perspective from like uh what I was saying is from the concentration perspective, but like uh, i'm also a software engineer. So from like software developing perspective， i'm also experiencing this. A new paradise shift is like, uh, previously I was a focused on computer region and when we were developing some a new a apps or maybe functions, we were thinking like, uh, where should we get the training data if if that actually like a real a real case happen like two months ago, like I was thinking having this uh uh developed this APP to uh maybe give you a score like on uh, how you look today but like a we were talking like from old perspective, I was thinking where should we get the training data? Like where should we get those daily look photos and get the scores but that all of sudden we will think you oh no, we we can have we can just directly use uh maybe GPT for all because like we are having this like common uh technology base so that we could uh use IT uh to do what we want uh previously is not uh is not possible but like now we are having this uh uh common technology base that we can use like we we have at our dispose .

yeah um I think with the help of A I the things that previously impossible, there are some uh demand supply relationship will be changed. So for example, take me as an example. I used to workless one or two a uh freeLancers outside uh all of our organization and right now I don't need to work with them because with the help of canvas and be journey I can do that myself so yeah i'm yeah is is is a pity that some people will lose their job. But I think this will change. Uh I think the A I has brought some change for the demand supply um so the next question will be， are there any changes with, uh, you get singing in the uh production relationship and content creation?

Uh maybe I will start first. Um is interesting when you mentioned that you use the doctor to uh, designers and you now don't need to talk to them. Maybe they are actually getting ten more clients because of ai.

So this is how we see IT because I think A I tools are especially productivity is actually enable user to do things in a much um faster way. So you know as any new technology um advancement rate, there will be some jobs that is being a made redundant. But at the same time, you. Release a lot of productivity um and that actually as a whole of the labor force action increase the productivity.

I think especially in in our industry were seeing that um uh fairly prominently because on one hand we we serve over formally and e commercials using our APP to create a content, create posters, create a collaterals to sell their stuff but you on the other hand we actually have um said cool, which is effectively the hands in china and uh on on that gooders over of seventy million a designers and artists on IT and their activity is you know also boom ming so were actually not seeing anybody suffered you know as a whole. Um so I think as as as a productivity um APP company or A I application company are our go with to really just try to see a Better tools so users can be more efficient. And at the end, I think the beneficiary will be consumers because you are able to get more details about the things that you want to buy. You get Better advertising, collect roles that at one point might be your personalized to you. So I think there will be your productivity roads along the way.

Yeah maybe I can add to that um from the more news ser information perspective um I would say there are two kind of phenomenon happening.

What is to what is like, if any any of you all if you think about how you access information of how you access news, the states right um it's very fragmented like we all get our information from different apps, different social media platforms from your friends um you know um on the I M S you know so it's no longer like ten years ago or twenty years ago where okay I I need this piece of news I go to your time. I need this piece of news I go to whatever media i've been following, right? That model has been falling apart.

So that actually combined with what we are talking about um selling, you use the word of civilized um I would say that democratically um of the content production side of things. So then now like um that comes with the information overload for all of us, right. Um i'm sure there must be a moment for all of us in the past week where you were just like overwhelming by the information, like he don't say nowhere to get that one piece information that you really need so on all these phenomena ding up that is actually giving what there are two chAllenges here.

What is the chAllenge for individual users? Um how do we actually um develop our own system to actually get to the information source or get to the piece of mention that we actually need right and that a chAllenge for the traditional kind of publishers on these providers instead, how do I meet this new demand from the audience and um and I think there is a bit of a dynamics happening. And one example is that now that would generate A I video content become more popular than before I compared to two years ago.

So um actually there has been a recent studies saying um because news content um assisted by Jennifer I has been becoming so popular, the old fashioned media companies and publishers are starting to produce more videos to meet this new demand. So I think there will be kind of more mutually dynamic um information ecosystem reshuffle happen in the next couple of years and we don't necessary know where that's going. Um but change is nothing happening.

Yeah I think the former put the formal analysis is great. Like what what I can say is just like my observation is like there are more and more contents pouring out and definitely there will be bad contents, not only the good quality, so, uh, that might be a problem, but uh, what I am a trying to a deal with this scenario is, uh, maybe just, uh, i'll i'm a lazy guy so i'll just like, let s go and I will pick the popular videos and I I believe the selection of like the audience right the audience will pick the the the good ones for me but alan, yeah yeah yeah algorithm and audience because like you you feed the audience like like meat to the algorithm, right? Yeah so like basically, that's still the choice of the audience.

So we talked about the innovation. Now IT comes to the problem, and we are going to talk about the what kind of new problems that they are created for us. I think news industry has been disrupted in the most significant way by the technology in the past ten to twenty years, especially given the development of A I generated content.

And there are some news about some, uh, new media outlet that using A I to generate their news and funded five users and readers. So I guess you learn have a lot of observations about this. So you longer can you give us some examples that you solve the new problems that credit by .

A I yeah definitely. Um so I think ultimately the news industry is a fact based industry, right? We we need to find the truth. We need to um develop the truth picture based on fact. Well, A I is not necessarily fact based.

If we know A I just enough, we know that that's actually based on statistics, right? So then it's generating the most probable whatever next word, next picture that based on the data IT has, which the data self is biased. Um mean many cases.

So then in that sense um that's actually I I mean I I create like the the first panel like that's great. I wish we can. We can embrace I like later, but we can not.

There are lots of journalist principles we have to stick to, which to some extent is not something that A I is able to do at the moment, right? Um so one one problem or chAllenges that actually come with this kind of a conflict is the the the overload of missing information line. And then this year is a big election year. Um there is a about, I think I might get the number wrong, but 4 yeah 4， thank you. Um forty different countries coming up with the elections this year.

So um billions of people are three elections, but they're also facing the chAllenges of missing information around elections and um regarding campaigns regarding different candidates um you know um I think that would say um example of that a real example of A A state in I think it's a wave in stating in the U S starting getting calls from joby den um and its audio course. So basically not so easy to make up like a ten minute an audio based on like ten seconds of someone's voice right um like we have A A I class uh A I coursing our school and the students can produce that within one class, like we can teach them how to basically producing this information. Um so but knowing that you know um they actually got the coffee, joe biden and telling them to to either vote for him or not to vote for a tram piano which end up as A A I generated audio message and not really um from biden of course so um this is coming not just a chAllenge for JoNathan m industry but also for broader society.

So that comes different chAllenges and different solutions we need to think of. You know, what about the media or information let me see of the audience like I think all of us, uh, pretty well educated, we probably know like how to identify the source. But what about you know people were less educ who may not have the the privilege to actually identify or think about people like our parents. My mom said misinformation all the time. So um how to actually do that kind of flitter acy training um that's something that um needs more ecosystem approach to address that.

Have this problems Sparked any discussion in the industry?

Yeah yeah I think as mentioned uh in my in my mark earlier um it's not just something that journals industry can address. Actually the reporters also um I think there are true gaps there.

One is reporting or enough discussion about A I accountability so you need to understand the problem enough to actually cover that and to actually tell the audience about IT right? So that's that's one discussion and practice start happening and the other uh may be touching a little bit on my U N part of work actually with that, one of the big projects we have been working on how to address election related missing information and how to have that ecosystem approach. Basically um we developed the product where we can actually use A I to do the fact checking process.

Um it's assisting it's not just like only using A I to do the fact chatting, but you can you say I to help you do some parts of the fact checking and to identify what are the more likely misinformation. And then you have humans following up. And then by that, we worked in developing countries and have the local media organizations to have the local networks and local languages to actually do the further step of human kind of wedding and human fact checking. So then you have the ecosystem media gania's A I slash technology and government and stepping in, and then the, you know, the general public. So then they at the end hopeful little can address um the the the problem we are facing .

altogether any input uh yeah .

uh maybe we can have uh a there is the same of a using magi C2Be mag ic. So the problem is caused by artificial maybe we can use artificial intelligence to have solved this problem, right? Like we can have this A I agent.

We were actually thinking about this. We can have this I agent to help class by misinformation and, uh, facts. And like we can, since it's based on A I, so IT can be really cheap and h everybody could have IT.

I can remember A M mark docker er said what what exactly you said before. A great do you have any input about this?

Um what I mean personally, I think this is actually not a new problem. I mean, depending on your perspective right now, fox news verses C N, one of one which is depending on your stances, misinformation so I think um as as I echo your this points um well education plays a very big part of IT and to a certain extent and it's easier to deserve what is A I generated verses you know human touch you know sort of this information rap in you know the principle under journalism at this point but obvious in the future.

Maybe it's it's more difficult, but I think things like this, what will happen anyway mean look back two hundred years, maybe one newspaper. Then newspaper means it's is the authority now you have fox first a CNN so so IT evolves over time. But I don't think this is um not solvable.

I mean it's a good chance for a new human as a whole to raise your critical thinking capability and try to design and with the help of you know magic to fight magic. I think this is also a way to do with that so much. He is still pretty optimistic about IT .

ah I think is not a war between human and artificial intelligence, always a war between human and human, like human with tools on to fight against .

each other yeah this is a chAllenge I think newsroom always facing um I I remember like ten or fifteen years ago in the first time I saw h five a new york times created my my instinct to us. This is a big chAllenge for newsrooms because for newsrooms they need a more and more engineers help them create those technological works to attract more users. But the question is, why do top engineers go to work at york times instead of facebook or google? So gary, have you been reached out by any uh contents, uh content craters or uh media groups to work with?

Um not I mean in what capacity?

I mean maybe just stay interested in talking with you guys to to see the potential of working list. Uh made me too jo.

Um I mean we we take regulate, uh interviews and all that, but in terms of using our tools to help the media agencies, I think they uh not particularly because because right now how how we organize a portfolio, have a further um industry specific uh strategy. So the first couple of APP is mostly focusing on the e commerce of industry.

Obviously as a media agency, you can still use our APP to create some of the new banners or advertised so off, but it's not gear to what's that. So I like you can use photoshop to edit your face too, but you are Better off using 美图秀秀。 Is that kind of analogy? It's easier that way? Yeah yes.

Uh so speaking of content creation, so this is the he like like what he said he was and he was study a machine learning when he was in his graduate school and he now is talking about A I his uh his his uh his video on 哔哩哔哩 is talking about lia。 So can you give us an example of how uh of the recent work you create IT with A I always the help of ai.

like my fight with A A I in terms right like uh like my contents。 Uh my my recent contents are all related to A I tools like we are using is to generate a illustrations and we use those silly straws in our videos and we I I definitely use uh ChatGPT or cloud to help with my uh process of writing my 嗯 articles。 And uh yeah that's that's basically my use case.

And the in in this process I was like, I I really struggle with like talking with A I agents. Uh, their responses is quick, their their attitude is good, but their capability is bad. I like, uh, what I think is a currently the A I agents, they like all those large language models. They have this problem of uh use steps is actually is not an original opinion from me, is actually from the the C E O of mini max.

I uh I borrows his idea like he he's I like the the current response of large language models are more like, uh in instinctive is like a what would you come to your mind like in in, in, in, in one second and it's not a deeply salt answer and when so that causes a very shallow use case. You can use, uh, A I agents. You can use large language models to do some simple tasks, but you cannot do complicate tasks like, uh, if I I prepare for this, uh for this like a for this session here, I have to go through many steps。

I need to like, look some, uh, look for some questions. I need to do some preparation. I need to like see all those documents and and generate some of my own ideas. But like for a large language models, that's just too much and so, uh basically, they are still doing some of the uh, additive work is like a supplemental work is not fundamental work. The work is still a through humans.

Just have any comments about this.

Yeah i'm kind of as I was saying, I was really thinking about, okay, how do I turn that into how I really educate the my students right? You know um some of them are are worrying about their their jobs. They're still studying. So they to start work, uh, looking for jobs seeing in next few month someone already worry about the jobs being gone. And I think what this is kind of describing is .

that you know if you .

are like you as human being, you just need to get smarter and even more systematic and strategic and thinking for you to actually be able to work with A I and have A I you in the ways that actually helpful, right um but to do that, you actually to really break down your tasks or you are thinking process in a very systematic way and only through that then um and I think that's for individual level but I think also for our industry level, right you need to break IT down so that you know how which part of the work flow can stepping and and support so I think that something all of us need to do more thinking about okay.

The time is limited so I will come to the the last part of this discussion and also a welcome questions from the audience um do .

we .

are all consumers and users of information too? So I wonder how we take control of what we read and what we consume the information every day. So do you have any advice or uh, for example, like where do you primarily get your information? And so is there anything you do deliberately to take control of the of the things you read and to tell which part A I has place in a piece of news?

So if I want to do some like formal study, I still go is like a formal journals and reports and papers. I'm working as a count creator. So I need to know what the community is talking about.

So sometimes I have to dive into that pull of garbage and find unlike what's our what's our audience talking about. So I I actually I I talked about that like previously。 Uh, what I did is just like I just uh lay back and see what the community or the the algorithm would they just feed me and and take a look and guess what people are talking about. And my suggestion would be like a as well as what i'm doing is actually I will have some time, maybe like uh one day or two a week to just stay away from like all of the contents .

in in general, I still stick to traditional news and all that um to to actually learn something very deep. I still prefer to know look at paper so talk to the expert by thing.

What A I comes in very handy is that before you talk to an expert or like a sea of another company or or professor, you know you you talk to GPT for for for a few moments first, and you can actually get you up to speed on some of the relevant information, give you A A little bit of concept so that you know you don't waste a lot of time when you are talking to the expert. And you know, granted, these larger english models will have some errors in IT, but that's where the expert comes in there here OK. Maybe if this is what you get from GPT, but then this is not that actually what this correct? So I think I can help again.

I think as human being, we have to find ways to let A I to increase the, but you're not trying to rely on A I to to sort of replace um human expertise. I think we are still not quite there yet. Don't know what's going going to happen in in ten years time. Read less well.

I'm glad to hear the old fashion journalists still have some business here. As for myself, I get most of information from social media but I think I do have the privilege of most of my social media networks are from like traditional journalists and media professionals. I kind of know that that's reliable information and instead of garbage um and one of my personal uh wish is actually I can build A I agent to manage my information intake. I don't know how to do that. If any of you have a have ideas or or where to start, i'm happy to get some inputs.

节目 的 最后 也 和 大家 分享 一则 我们的 生动 胡同 会员 的 一个 消息。 九月份 正值 开学 季， 我们 希望能够 让 我们的 学生 朋友们 更加 便宜 的 成为 我们的 会员。 所以 从 去年 开始， 我们 就 增加 了 一个 看起来 不怎么 合理， 但是 大家 都 很爱 的 一个 教育 折扣。 正常 加入 我们 互通 会员 其实 是 365元1年。 如果 使用 我们的 教育 折扣， 其实 是 120元， 其实 是 将近 3折的 这样 一个 优惠 力度。 成为 我们的 会员 可以 解锁 我们 旗下 所有的 付费 节目， 包括 不值 金钱， 还有 跳进 兔子洞。 同时 还 可以 收到 我们 每 周一封 的 邮件 通讯， 就是 news letter 里面 记录 了 我们 制作 内容 的 幕后 故事， 然后 以及 我们 当下 的 所思 所 想。 点 开 我们的 说 no， 也就是 节目 详情 中的 文章 链接， 可以 看到 我们 为什么 设计 这样的 一个 特别 有 力度 的 一个 折扣 给 我们的 学生 小伙伴 们。 然后 也会 有 我们 生动活泼 的 小伙伴 们 对 自己 18岁说 的 一句话， 欢迎 各位 来 订阅， 也 欢迎 大家 把 这样的 一个 特别 有 力度 的 折扣 分享 给 身边 正在 上学 的 小伙伴 们。