OpenAI 用 ChatGPT 构建自定义预测应用
Use ChatGPT Work to build custom forecasting apps
This has been a really awesome way to do forecasting. It saves a lot of time to do more strategic forward-looking analysis or uh initiatives for our finance team. Hey everyone, I'm Jackson. Um I'm on the data science team here at OpenAI, but I work really really closely with the strategic finance team. Um and it's really exciting to talk to you a little bit about how we do forecasting here at OpenAI. all the folks on the call on the finance teams every single um month you can imagine the beginning of the month scramble where you're waiting for the data to land you're updating your spreadsheets you're trying to talk to the product team the marketing team the go to market team of like how you should be updating your forecast and you're trying to build a really solid forecast that you can present to the CFO I think everyone kind of feels that scramble um it's no different here at OpenAI um but we're always asking ourselves how can we make this process better so the the problem we wanted to solve was to build something like an interactive UI that we could send over to our finance partners, work with them collaboratively and ultimately update our forecasts.
这真是一种非常棒的预测方式。它节省了大量时间，让我们财务团队能够进行更具战略性的前瞻性分析或举措。大家好，我是Jackson。嗯，我是OpenAI数据科学团队的成员，但我与战略财务团队合作非常非常紧密。嗯，很高兴能和大家聊聊我们在OpenAI是如何进行预测的。所有参加电话会议的财务团队成员，你们可以想象每个月月初的忙碌：等待数据落地，更新电子表格，试图与产品团队、市场团队、市场推广团队沟通，了解应该如何更新预测，并努力构建一个可靠的预测，以便向CFO汇报。我想每个人都能感受到那种忙碌，在OpenAI也不例外。但我们一直在问自己，如何能让这个过程变得更好。所以，我们想解决的问题是构建一个交互式UI，可以发送给我们的财务合作伙伴，与他们协作，最终更新我们的预测。
The first thing I did was I'm prompting I'm prompting chat to say, hey, let's kick off this monthly forecast. Um, pull the latest actuals, update our data science baseline, and build an interactive finance forecast using sites. sites is basically a way for anyone to kind of build a website with no code, host it without any infrastructure and it's something we use a lot internally and let's see what it came up with. We see the data science baseline, we see our prior forecast and we see kind of a scenario we can build and this is all at the aggregate level right here.
我做的第一件事是提示Chat，说：嘿，让我们开始这个月的预测吧。嗯，拉取最新的实际数据，更新我们的数据科学基线，并使用sites构建一个交互式财务预测。sites基本上是一种让任何人都能无需代码构建网站、无需基础设施即可托管的方式，这是我们在内部经常使用的东西。让我们看看它生成了什么。我们看到了数据科学基线，看到了之前的预测，也看到了我们可以构建的情景，而这一切都在汇总层面。
Um, I have a visual representation of this as well, so I can kind of see if the trends are looking like, you know, eyeball the trends to see what we're what we're thinking for the next uh half of the year. And then um I've also asked it to look account by account so I can make kind of adjustments by account. And um this this site has a feature to go down every account, but let's look at Apex, one of our larger accounts.
嗯，我也有一个可视化的表示，所以我可以大致看看趋势，比如，观察趋势，看看我们对下半年有什么预期。然后，我还要求它按账户逐一查看，这样我就可以按账户进行调整。嗯，这个网站有功能可以下钻到每个账户，但让我们看看Apex，我们较大的账户之一。
So we can kind of see the baseline here. And then we can kind of see this green line here that we can actually adjust directly in the UI which is super cool. And what's happening right now is it's although it's updating the forecast for this account the entire model is updating in the background. So the aggregate number also gets adjusted.
所以我们在这里可以看到基线。然后我们可以看到这条绿线，我们可以直接在界面中调整它，这非常酷。现在发生的是，虽然它正在更新这个账户的预测，但整个模型在后台也在更新。所以汇总数字也会被调整。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力