这周周末活动多，导致没来得及写周报，现在才补上

先说说这周看到的有趣链接

一个转为developers打造的GPT搜索引擎

给我的感觉是可以用，代码什么的可能比原本的好一点）可能就是提示词吧

这是hackernews对它的讨论

这是一个浏览器插件能时刻使用上bing

我的bing也开始跳转国内了，这个插件还不错，能正确显示代码块

此外我很乐意展示我使用bing的一次示例，帮我完成课内代码：

一次new逼使用纪实

user: can you read this code?

squares algorithm

phiN = [y(3:1e6-1,1),y(2:1e6-2,1),y(1:1e6-3,1),u(3:1e6-1,1),u(2:1e6-2,1),u(1:1e6-3,1)];
yN = [y(4:1e6)];
disp("least squares algorithm");
theta = (phiN'*phiN)\phiN'*yN

bing: Yes, I can read this code. It is a MATLAB code that implements the least squares algorithm to estimate the parameters of a linear system. The code does the following steps:

It defines a matrix phiN that contains the past values of the output y and the input u of the system. The matrix has 999997 rows and 6 columns.
It defines a vector yN that contains the current values of the output y of the system. The vector has 999997 elements.
It displays the message “least squares algorithm” on the screen.
It calculates the vector theta that contains the estimated parameters of the system. The vector has 6 elements. The calculation uses the formula theta = (phiN’*phiN)\phiN’yN, which is derived from minimizing the sum of squared errors between yN and phiNtheta.

user: please rewrite it by using Recursive Least-Squares Algorithm

bing: [1]: https://www.mathworks.com/help/ident/ref/recursiveleastsquaresestimator.html “Estimate model coefficients using recursive least squares (RLS …”
[2]: https://www.mathworks.com/matlabcentral/fileexchange/56360-recursive-least-squares “Recursive Least Squares - File Exchange - MATLAB Central - MathWorks”
[3]: http://freesourcecode.net/matlabprojects/59802/rls-matlab-code “Rls Matlab Code | download free open source Matlab toolbox, matlab code …”
[4]: https://ai.stackexchange.com/questions/5715/can-we-use-the-recursive-least-squares-as-a-learning-algorithm-to-an-adaline “Can we use the recursive least squares as a learning algorithm to an …”
[5]: https://www.eit.lth.se/fileadmin/eit/courses/ett042/CE/CE5e.pdf “Computer exercise 5: Recursive Least Squares (RLS)”

The Recursive Least-Squares (RLS) algorithm is a method to estimate the parameters of a system that is linear in those parameters, by minimizing a weighted sum of squared errors. The RLS algorithm updates the parameter estimates recursively as new data arrives, and uses a forgetting factor to give less weight to old data.

To rewrite the code by using RLS algorithm, you can use the following steps:

Define the forgetting factor lambda and the initial values of the parameter vector theta, the estimation error covariance matrix P, and the regressor vector phi.
Initialize a loop to iterate over the data points from 4 to 1e6.
Update phi with the current values of y and u.
Update theta and P using the RLS formulae.
End the loop.

The rewritten code is:

% Recursive Least-Squares algorithm
lambda = 0.99; % forgetting factor
theta = zeros(6,1); % initial parameter vector
P = eye(6); % initial error covariance matrix
phi = zeros(6,1); % initial regressor vector
for k = 4:1e6
    phi = [y(k-1); y(k-2); y(k-3); u(k-1); u(k-2); u(k-3)]; % update regressor vector
    g = P*phi/(lambda + phi'*P*phi); % gain vector
    theta = theta + g*(y(k) - phi'*theta); % update parameter vector
    P = (P - g*phi'*P)/lambda; % update error covariance matrix
end
disp('Recursive Least-Squares algorithm');
theta % display final parameter vector

user: could give me a RLS code without forgetting factor lambda

bing: [1]: https://bing.com/search?q=matlab+rls+without+forgetting+factor “Compute forgetting factor required for streaming input data - MATLAB …”
[2]: https://www.mathworks.com/help/dsp/ref/rlsfilter.html “Compute filtered output, filter error, and filter weights for given …”
[3]: https://www.mathworks.com/help/fixedpoint/ref/fixed.forgettingfactor.html “Compute forgetting factor required for streaming input data - MATLAB …”
[4]: https://www.researchgate.net/publication/251910252_A_variable_forgetting_factor_RLS_adaptive_filtering_algorithm “A variable forgetting factor RLS adaptive filtering algorithm”

To give a RLS code without forgetting factor lambda, you can simply set lambda to 1 in the previous code. This means that the algorithm does not forget any past data and gives equal weight to all data points. The code is:

% Recursive Least-Squares algorithm without forgetting factor
lambda = 1; % no forgetting factor
theta = zeros(6,1); % initial parameter vector
P = eye(6); % initial error covariance matrix
phi = zeros(6,1); % initial regressor vector
for k = 4:1e6
    phi = [y(k-1); y(k-2); y(k-3); u(k-1); u(k-2); u(k-3)]; % update regressor vector
    g = P*phi/(lambda + phi'*P*phi); % gain vector
    theta = theta + g*(y(k) - phi'*theta); % update parameter vector
    P = (P - g*phi'*P)/lambda; % update error covariance matrix
end
disp('Recursive Least-Squares algorithm without forgetting factor');
theta % display final parameter vector

一些不知疲惫的“制作无限曲别针”的GPT

2003年人类对强人工智能提出了一个曲别针制造机假说，提出制造曲别针这个无害的指令也会对人类造成不可挽回的后果。

最近有很多类似的项目出来了

最初的babyGPT 这里可能是对应代码，我没细看

然后出现了 autogpt

这个玩意儿短短几天就102kstar了，比比特币疯狂多了

他相当于第一次打开了制造曲别针的这个潘多拉魔盒

以及从autoGPT衍生出来的agentGPT,这是个网页版本，可以填上api直接跑，虽然效果很傻

colab教程

让我在使用colab上有迹可依

因为睡午觉而错过了4月20号北京首钢主场球票开售的我，不得不经历一系列问题，于大麦网抢票打交道，因此萌生了做抢票脚本的想法。但是网上没有开源的东西，毕竟这种东西对黄牛来说违法又赚钱，只能自己想办法以后多学学了。
这个xqc让我短暂萌生了做tts的想法，后来调查了一下发现人家帮他做这个东西的粉丝是个真的大牛，遂感叹自己弱鸡

视频链接，如果上面视频不能播放的话）

这里是大牛的推特