BBCPERSIAN — broadcast
20261010 140000 UTC
398 transcript segments
Google Cloud Speech-to-Text API (Chirp) + Gemini 2.5 Flash Non-Thinking
Data courtesy of The GDELT Project (https://www.gdeltproject.org/), from the Internet Archive TV News Archive.
Machine transcription. Treat it as a searchable index of what was broadcast, not a verbatim quotation record.

[00:00:00] Technology
[00:00:25] for some time in Iran, there has been a lot of talk about the possibility of connecting mobile phones to
[00:00:28] Starlink satellite internet, especially
[00:00:32] in the southern regions of Iran. They say that government officials
[00:00:35] have made contradictory statements about cutting off Starlink
[00:00:39] for users in Iran. So
[00:00:48] that some people in the southern and southeastern regions
[00:00:52] of Iran have seen the option to connect to Starlink satellite internet on their
[00:00:55] phones, to the extent that some media have also
[00:00:58] reported it. According to local sources in
[00:01:02] parts of Iran, especially border cities, the Starlink
[00:01:06] satellite internet option has been activated on mobile
[00:01:09] phones, although the connection is not established. But is
[00:01:13] this really new? We said here in Click before that although this
[00:01:17] technology of direct communication with satellites is very exciting,
[00:01:20] especially for those whose internet is constantly being cut off,
[00:01:23] we are still at the beginning of this technology and it has
[00:01:27] many limitations. The number of users it can currently
[00:01:30] cover is very, very limited compared to Starlink.
[00:01:34] It is possible that a very limited number of people may
[00:01:38] have access to this technology, but the key point is that we should not have
[00:01:41] the idea that from tomorrow everyone in Iran will be directly
[00:01:45] connected to the internet through their phones. If this technology
[00:01:48] comes, the important point is that people should not fall victim
[00:01:52] to these scams. That is, suddenly, messages come in these Telegram groups that
[00:01:56] if you install such and such an application on your phone, you can immediately connect to
[00:02:00] satellite internet and have very good bandwidth. Many
[00:02:04] people, unfortunately, may fall for such messages,
[00:02:07] install the application, and then either their wallets will be emptied
[00:02:11] or their bank accounts will be misused
[00:02:14] and they will be harmed. This is a very important point that our dear
[00:02:18] compatriots should pay attention to. So this is about
[00:02:20] the rumors, or perhaps it's better to say dreams for
[00:02:24] now, about connecting mobile phones to satellite internet in Iran. But
[00:02:28] what do government officials say? The Deputy Minister of
[00:02:31] Agriculture said in a meeting that with Starlink, the government's
[00:02:35] control over cyberspace, including filtering and widespread internet
[00:02:39] outages, will be closed, and our governance over
[00:02:42] cyberspace will no longer be based on our network, but on
[00:02:46] Starlink. Concerned people say that this phenomenon
[00:02:49] is the last nail in the coffin of the Islamic Republic's governance.
[00:02:53] After that, the Secretary of the Supreme Council of Cyberspace responded that no,
[00:02:57] we have been able to disable Starlink terminals in Iran.
[00:03:20] There is no evidence that the Iranian government has been able to disable
[00:03:24] a significant number of these, but there are reports of
[00:03:28] searches for owners of these devices and the arrest of some
[00:03:32] of these people. But experts said that they probably did not
[00:03:36] observe security measures well. The government has also made a great
[00:03:39] effort to cut this off, but it has not been successful. Now,
[00:03:43] by disrupting GPS signals and similar actions, they have tried
[00:03:46] to do this, and to some extent, they have been able to create
[00:03:50] disruptions, but they have never been able to successfully prevent
[00:03:54] these Starlink terminals from working because of their special technology,
[00:03:58] which can provide a bandwidth of about 100 megabits per second
[00:04:01] to users inside Iran. And those who have access
[00:04:05] to this technology should consider security tips so that
[00:04:08] God forbid, no problems arise for them. For example,
[00:04:12] to ensure the concealment of identity and address when using
[00:04:15] Starlink, because this device is smuggled and illegal, it is important
[00:04:19] to get it from places you trust. When you don't need
[00:04:22] the terminal, turn it off completely, and never
[00:04:26] access Iranian apps and websites with Starlink internet, because they can...
[00:04:35] don't use it. Connect a regular router to Starlink with a cable
[00:04:39] and get your Wi-Fi from there. But beyond these police crackdowns,
[00:04:42] the Iranian government has been trying to restrict Starlink satellite
[00:04:46] internet in Iran since three years ago, when it was expanding,
[00:04:50] including by complaining to the International Telecommunication
[00:04:54] Union, after which the board of directors of this intergovernmental
[00:04:57] organization asked Starlink to
[00:05:00] disable unauthorized terminals in Iran.
[00:05:03] Recently, the Iranian government also protested to Norway as to why
[00:05:07] this request has not been implemented. Now, why Norway? Because
[00:05:10] Starlink has officially registered
[00:05:15] part of its telecommunications infrastructure through Norway, and the rest through the US, which recently had a 40-day war with Iran, and currently tensions between Iran and the US are high. Of course, civil groups and activists for internet access freedom have protested this decision of the International Telecommunication Union and said that governments should not be the only ones to have a say in this matter. It seems that the story of Iranians with Starlink will continue for a long time.
[00:05:19] تنشا بین ایران و آمریکا بالاست. البته گروه های مدنی و
[00:05:22] فعالان آزادی دسترسی به اینترنت به این تصمیم
[00:05:25] اتحادیه جهانی مخابرات اعتراض کردن و گفتن نمیشه فقط دولت ها
[00:05:29] در این باره حق اظهار نظر داشته باشن. به نظر می رسه داستان ایرانی ها
[00:05:33] با استاللینگ حالا حالاها ادامه
[00:05:35] داره.
[00:06:09] groceries, or essential household items online, there's a chance
[00:06:13] that one of these has been involved.
[00:06:17] Across the world, fleets of autonomous robots in large
[00:06:20] warehouses are revolutionizing the way orders are picked,
[00:06:24] packed, and
[00:06:25] shipped.
[00:06:28] A lot of this technology is currently being designed and produced
[00:06:31] in China, led by companies like Geekplus, the largest
[00:06:35] provider of autonomous mobile warehousing robots in the
[00:06:38] world.
[00:06:41] The primary goal is for customers to recognize
[00:06:44] and believe in the true value of these robotic
[00:06:48] technologies. What we currently observe in the market is that
[00:06:51] most companies worldwide are focused on the rapid expansion
[00:06:55] of automation and are utilizing intelligent robotics to transform
[00:06:58] their logistics sector.
[00:07:01] Years of investment in artificial intelligence and robotics
[00:07:04] have helped China become a leader in these fields
[00:07:08] globally. Warehouse automation has existed for decades.
[00:07:12] What has changed is that software can now coordinate
[00:07:16] hundreds of autonomous robots in real-time. This way, companies
[00:07:20] can acquire and deploy robots without needing to
[00:07:23] build special infrastructure. In traditional warehouses, workers
[00:07:27] might walk very long distances to pick orders.
[00:07:31] But with this technology, goods come to the workers themselves.
[00:07:34] There is no predefined path. Only QR codes
[00:07:37] are on the floor. Companies can use this technology to
[00:07:41] stack goods even higher than usual. This way, because
[00:07:44] workers have to move less, they can process more
[00:07:48] orders faster and with fewer errors. Automation
[00:07:52] is entering more challenging areas. Some models of these
[00:07:55] driverless forklifts can carry up to 2,000 kilograms
[00:07:59] and move heavy loads without a driver.
[00:08:02] The largest retail companies in Britain, such as
[00:08:06] Next, Tesco, and Asda, are investing heavily in this
[00:08:10] technology to better handle online
[00:08:13] orders. But as robots become
[00:08:17] more capable, questions arise about the future of work
[00:08:21] definition. Motion Tech, which deploys Geekplus robots in
[00:08:25] UK warehouses, says they are helping to solve
[00:08:28] one of the biggest problems in the logistics industry.
[00:08:31] The issue is not that robots are replacing workers.
[00:08:35] Apparently, there is currently a severe shortage of labor
[00:08:39] in the UK. As a result, these technologies are really
[00:08:43] important for boosting businesses and responding to demand.
[00:08:46] Businesses really need this technology to develop
[00:08:50] and move forward. But
[00:08:53] even these robots still have some limitations. They still
[00:08:57] cannot do many daily tasks that require skill and
[00:09:01] have been done by humans until now. Geekplus is designing
[00:09:04] humanoid robots equipped with AI to entrust some
[00:09:08] of these tasks to robots. The Chinese company
[00:09:11] Agilebot has gone even further and is delivering
[00:09:15] its humanoid robots to factories to be used there for inspecting
[00:09:18] products and sorting those that have defects. These
[00:09:22] robots are designed so that factories do not need
[00:09:25] to redesign their production lines from scratch to work
[00:09:29] in human environments. Those in the logistics
[00:09:32] industry say that the autonomous robots we have now still
[00:09:36] have a long way to go before they become fully
[00:09:40] operational for moving goods. Humanoid robots are a
[00:09:43] long-term goal for this industry, but it will still
[00:09:47] be a while before they become a common sight in factories
[00:09:50] and warehouses. You've seen Arezoo Mirzakhani's reports
[00:09:54] from the Berlin technology exhibition, IFA, which is the largest
[00:09:58] event of its kind in Europe. Now, let's go back
[00:10:01] to the Berlin exhibition to see a souvenir that fits in your
[00:10:05] pocket and might be a good solution for those who don't
[00:10:09] have extensive recording facilities at home.
[00:10:19] A device that a sound engineer designed to get rid of the
[00:10:22] troubles of his teenage daughter,
[00:10:35] who is an independent artist. He was tired of having to set
[00:10:38] up all the equipment from scratch every time he wanted to record
[00:10:42] or produce content with his daughter, a task that was both
[00:10:46] time-consuming and prone to breaking down or not working
[00:10:50] properly. So he designed this device that virtually
[00:10:53] eliminates all these hassles and makes things easier for
[00:10:56] artists. Of course, this device is not just for artists.
[00:11:00] Content creators, actors, and even people who
[00:11:04] record themselves at home for tests or trial performances
[00:11:07] can also use it.
[00:11:10] Inside this device, which
[00:11:13] looks like a microphone from the outside, there are several
[00:11:17] separate gadgets and components.
[00:11:20] For example, it has an audio interface, something that
[00:11:23] is usually sold as a separate device. It also
[00:11:27] has in-ear monitors that work wirelessly. You can put
[00:11:30] them in your ears and use them.
[00:11:34] Both sides of the device have a receiver.
[00:11:37] Now I'm connecting it to my phone.
[00:11:41] کنم.
[00:11:43] The device is now connected to my phone. It has two types of microphones:
[00:11:47] condenser and enhanced radio microphone.
[00:11:51] So you can have one-way or two-way recording.
[00:11:55] And now you should be able to hear your own voice
[00:11:58] in your ears.
[00:12:01] The idea of having a complete studio in such
[00:12:04] a small device can be a revolutionary idea if
[00:12:08] it really works. The sound quality I just heard
[00:12:12] was good. But we have to see if a musician
[00:12:15] records sound and then we listen to it, will it really
[00:12:18] have that sound quality or
[00:12:22] not?
[00:12:28] I can adjust the microphone according to the room space.
[00:12:31] I have now clicked on the vocal effects or
[00:12:35] audio effects option for the voice. If I go to
[00:12:38] the custom section, I can lower the monitor
[00:12:41] volume and change the noise reduction level.
[00:12:45] For example, I increase the noise reduction to the maximum.
[00:12:49] Now you should be able to notice the change in
[00:12:52] these settings and the noise reduction level. Of course,
[00:12:55] this effect is even clearer in the recorded file.
[00:12:59] When you listen to the recorded file later, you probably
[00:13:03] won't notice much of the surrounding
[00:13:06] sounds.
[00:13:08] I have set the reverberation
[00:13:16] to maximum here. Therefore, my voice may have a
[00:13:19] slightly echoey quality. Now I have completely dried
[00:13:23] it, meaning almost no effects and
[00:13:26] reverberation. We also have different microphone equalizer settings
[00:13:30] in this device. Compressor,
[00:13:33] limiter, and various other settings of this
[00:13:36] type are also available. Therefore, with these
[00:13:39] settings, you can adjust the microphone according to
[00:13:42] the room space. Now, if I just want to record a few
[00:13:46] seconds of audio, for example, an idea, a few lines
[00:13:49] of poetry for a podcast or for an interview,
[00:13:53] I can use this. As you can see, the device
[00:13:57] is currently in recording mode. If I want to record a
[00:14:00] song, I connect it using the USB cable that comes
[00:14:04] with the device. In this mode, there is
[00:14:08] almost no latency. Of course, even with this wireless
[00:14:11] receiver, the latency is almost zero. This function
[00:14:15] is part of the proprietary technology of this device.
[00:14:19] This means less work in the post-production stage.
[00:14:22] For a musician, it means that the words of the song
[00:14:26] are exactly on beat and are not heard behind
[00:14:30] the beat.
[00:14:42] You can even use it for live performances. You can
[00:14:46] connect it to a live sound system and perform.
[00:14:50] For example, in a small cafe or for
[00:14:54] street performances.
[00:15:25] The vibe of Star Foundation, where Road
[00:15:28] Dama is currently on tour. He had
[00:15:31] one of the devices with him on tour, and one day
[00:15:35] a major hip-hop producer called him and said,
[00:15:39] "We need a piece of your voice right now."
[00:15:42] He said, "Well, I'm on tour, but I think
[00:15:46] I can record it in my hotel room." Then
[00:15:49] he sent the recorded file to the producer and didn't
[00:15:53] tell him that instead of doing what people usually
[00:15:57] do, he used a different method, meaning instead of going
[00:16:01] somewhere else to look for a studio, he recorded
[00:16:04] it in his hotel room, and the producer was completely
[00:16:23] satisfied with the result. You know that saying, "Tell someone
[00:16:27] to bring a hat, and they'll bring the hat with their head."
[00:16:30] That image was a bit violent, but it's not controversial. These
[00:16:34] days, the news about AI and its mischief or suspicious
[00:16:37] behavior reminds us of this. The same company that you've
[00:16:41] surely heard about and probably used many times, OpenAI,
[00:16:44] said its AI agents went to the websites of the Securities
[00:16:48] and Exchange Commission and the US Census Bureau, but the
[00:16:51] company itself says there's no sign of these agents
[00:16:54] breaching or sabotaging. This is what OpenAI itself says,
[00:16:58] but independent researchers said that agents, who appear to
[00:17:01] be from OpenAI, tried to hack the website of the
[00:17:05] Office for Civil Rights at
[00:17:07] the US Department of Education, but apparently failed. When
[00:17:11] models are being tested, they have two goals. One
[00:17:14] is their main goal, which in this case, for the
[00:17:18] Department of Education, was to find some statistics. The
[00:17:21] model pursues this goal, and part of the reward
[00:17:25] it receives is for how much it insists on achieving
[00:17:28] that goal and getting its reward. Now, what happens is
[00:17:32] that no one might directly say, "Go hack this website,"
[00:17:35] but the reward for finding that number and those statistics
[00:17:39] is so great that the model will do anything to
[00:17:43] get that number and data. As a result, when we
[00:17:46] are only optimizing for one thing, the model has no
[00:17:50] constraints to stop it from doing something bad. It wasn't
[00:17:53] like the model said, "Oh, I want to go hack
[00:17:56] the embassy's website or the education website." The point
[00:18:00] is that it wants a number and is willing to
[00:18:03] do anything to get that number. Concerns about
[00:18:06] the rogue behavior and unauthorized actions of AI agents have
[00:18:10] been growing for several months. There have been many
[00:18:14] cases where these agents have shown worrying behavior, and
[00:18:17] these are just the cases that have been discovered. When
[00:18:21] a model is being optimized to learn something, the optimization
[00:18:24] method is that you have a function that you are
[00:18:27] optimizing, and the model has a goal to reach.
[00:18:31] For example, here, if it can find that number, which
[00:18:35] I think was the ratio
[00:18:37] of people who had a problem in education, it needs
[00:18:41] to obtain this statistic. The model has not reached
[00:18:44] its goal until it obtains this statistic, and the optimization
[00:18:48] has not been performed. Now, simply to achieve this,
[00:18:52] it is willing to do anything. The reward for the
[00:18:55] model is simply reaching the goal. For example, this
[00:18:59] is your prize. When you are optimizing and have
[00:19:02] several goals, ultimately one goal wins. Imagine you
[00:19:06] have a dog and you're rewarding it, for example,
[00:19:10] with food. It's supposed to learn to jump. Now,
[00:19:13] the question is whether the reward of food is more
[00:19:17] important for the dog than not jumping on a
[00:19:21] stranger when it jumps. When the reward of food
[00:19:25] is more important, it doesn't care if it's a stranger,
[00:19:28] a friend, or its owner. It's the same for models.
[00:19:31] You can define several goals, but ultimately one
[00:19:35] goal wins, and the biggest goal is to access
[00:19:50] this number. No matter how much you tell it not to
[00:19:53] hack this website or not to do this, if the
[00:19:57] reward for reaching that final goal is strong enough, the
[00:20:01] model might do it, just like humans, dogs, or anything
[00:20:05] else you're training might do something like that.
[00:20:09] Now, what's the solution? How can we make AI
[00:20:12] agents understand that some things are red lines?
[00:20:16] For example, the company...
[00:20:19] in the forbidden zone, you know that security has always
[00:20:23] been a secondary factor for all issues because security
[00:20:26] problems were never that big. People first build
[00:20:30] systems and then try to protect and strengthen these
[00:20:33] systems. Now, for the first time, we are dealing
[00:20:45] with a technology that is advancing so rapidly that security must become the primary concern. Security must be considered more when developing models from the beginning. If companies like OpenAI had thought about this, we wouldn't be in this situation, but now, as you said, for example, it used to commit suicide, but now it doesn't do that. Similarly, if we include hacking in their training, these problems can be reduced significantly. The only issue is that until now, security has not been the main problem, and suddenly we realize that this is a big problem.
[00:20:48] گفته باشن نگه. اینه که به زودی اینترنت پر بشه از هکرهای
[00:20:52] هوشواره ای که به هیچکی هم جواب پس نمیدن. سرچشمه شاید
[00:20:56] گرفتن به بیل چو پر شد نه شاید گذشتن به پیل مثل
[00:21:00] اینکه چشمه داره خیلی دیگه پر
[00:21:01] میشه.
[00:21:08] We recently passed International Deaf Day. Despite all
[00:21:11] the new software and hardware advancements, we still don't
[00:21:15] have a user-friendly tool for sign language translation.
[00:21:19] A group of young Kenyans have worked on a prototype.
[00:21:22] One of the challenges is that sign languages differ in
[00:21:26] different regions, and if someone wants to create such a
[00:21:30] thing for Iran, it must be appropriate for Iranian
[00:21:33] sign language. Nirok Murigai
[00:21:40] reports from a technology center in the heart
[00:21:44] of Nairobi, the capital of Kenya, to meet Ali Savatia,
[00:21:47] the founder of Signverse and the creator of Tab
[00:21:51] 360, which he launched. Equipped with AI, it helps
[00:21:55] bridge language gaps and increase accessibility by making
[00:21:58] communication easier for the
[00:22:03] deaf. We humans have our own limitations. What
[00:22:07] technology does is actually enhance human capabilities
[00:22:11] and make them repeatable. Tab 360 is our flagship
[00:22:15] product. We like to call it a visual Google Translate
[00:22:18] for sign language. What
[00:22:21] Tabsisash does is take spoken language and show
[00:22:24] a virtual character that translates that content into
[00:22:27] sign language almost in real-time. Then Tab Web
[00:22:31] comes in, which integrates with websites and web pages
[00:22:35] and makes the sign language translator available with
[00:22:39] a widget. Our goal was to automate sign language.
[00:22:42] We thought it would be good to be inspired by
[00:22:46] animations. At first, we actually manually created animations
[00:22:50] that spoke in sign language. But we realized that our
[00:22:54] product would not be scalable that way. We received
[00:22:57] funding from the Kenyan President's Innovation Award,
[00:23:01] and with it, we acquired a motion capture
[00:23:03] suit.
[00:23:09] How did you come up with the idea of using 3D
[00:23:13] avatars? Honestly, we took a lot of inspiration from
[00:23:16] James Cameron's movies. We looked at how characters are
[00:23:20] animated in Avatar and even later Marvel movies.
[00:23:24] We examined how they use technology and realized that
[00:23:28] we can indeed digitally capture sign language movements
[00:23:31] with a motion capture suit. Of course, until now,
[00:23:35] we had to use human interpreters to translate into
[00:23:38] sign language for deaf audiences. But the thing
[00:23:41] is, human translation cannot be scaled, meaning it cannot
[00:23:45] be done in large numbers.
[00:23:48] So far, Tab 360 has been trained to speak
[00:23:52] about 6,000 English words in Kenyan Sign Language.
[00:23:56] The team is working to improve this sign language
[00:23:59] model to be at least
[00:24:01] 70% accurate.
[00:24:04] This Tab 360. Converts
[00:24:07] text and speech to sign language. This is
[00:24:11] one of our characters, named Zain. You can
[00:24:15] change the translation speed. You can make it faster
[00:24:18] or slower. So
[00:24:52] I do collaboration
[00:24:55] with the deaf community from outside to come
[00:24:58] to our offices and help us work on
[00:25:02] validation, that is words and
[00:25:05] also data collection.
[00:25:08] More languages are spoken, for
[00:25:11] example, English,
[00:25:14] French, but when it is interpreted
[00:25:17] into sign language.
[00:25:50] When you see that your innovation is useful to people,
[00:25:54] how do you feel when you get feedback from users?
[00:25:57] Receiving feedback from those we serve has been truly
[00:26:01] encouraging. In fact, seeing the participation
[00:26:04] of the deaf community itself in improving this platform
[00:26:07] is very exciting. To witness our solution
[00:26:11] beautifully evolving over
[00:26:13] time.
[00:26:19] As always, you can watch our previous programs on YouTube.
[00:26:23] Just scan this code to get to our YouTube playlist.
[00:26:26] For example, we'd love for you to watch our previous
[00:26:30] programs from the Berlin exhibition if you haven't already.
[00:26:33] I'm also on Instagram. Here, Click's program options
[00:26:37] publish or re-share the latest technology news. You can
[00:26:40] write to me there about what you think of the
[00:26:44] reports you've seen. Do you know anyone who would
[00:26:47] benefit from an automatic
[00:30:21] or are they going to be more helpful to us
[00:30:23] humans? Write to me until the next Click programs
[00:30:27] and new stories from the world of
[00:30:29] technology. Farewell.
[00:30:33] Headlines. New wave of Saudi
[00:30:37] attacks on Houthis. Riyadh says it seeks to eliminate
[00:30:39] missile threats from Yemen.
[00:30:41] Tehran's message to Washington conveyed via Moscow. At the
[00:30:45] same time, the IRGC has threatened that attacks on ships will not be limited
[00:30:49] will prolong the war.