AI just created a brand new virus. Should we be scared? | BBC News — Transcript

Full transcript

1. 0:00So, here we go again with what could be
2. 0:01a plot in a sci-fi thriller, but is
3. 0:04reality.
4. 0:05>> Researchers in the US have for the first
5. 0:07time [music] used artificial
6. 0:08intelligence to design new viruses.
7. 0:10>> It's been hailed a very significant
8. 0:12turning point. What could possibly go
9. 0:14wrong?
10. 0:14>> There's also concern that the technology
11. 0:16could create new
12. 0:18>> [music]
13. 0:18>> and yet unheard of bio weapons.
14. 0:26>> It does sound like something from a
15. 0:28sci-fi movie, but this is real. A US
16. 0:31team have used artificial intelligence
17. 0:33to design completely new viruses. Well,
18. 0:36coming up, we'll assess what it means,
19. 0:38including questions it raises about
20. 0:41safety and biosecurity. First, what
21. 0:44exactly did the researchers actually do?
22. 0:47>> AI has been used to design simple
23. 0:50structures like antibiotics before.
24. 0:52>> But now, for the first time, AI has
25. 0:54designed the entire genetic code, the
26. 0:56whole genome of viruses that aren't
27. 0:59found in nature.
28. 1:00>> Scientists took these AI designs and
29. 1:03attempted to make them in a lab.
30. 1:05>> It works a bit like chat GPT, but
31. 1:08instead of predicting sequences of
32. 1:10words, these AI models predict the
33. 1:12language of life.
34. 1:14>> And instead of being trained on books
35. 1:15and websites, these AI models were
36. 1:18trained on genetic codes from viruses,
37. 1:21bacteria, plants, and people.
38. 1:23>> They were then refined to create a type
39. 1:25of virus known as a bacterial [music]
40. 1:28phage, which infect only specific
41. 1:30species of bacteria.
42. 1:32>> 16 virus designs were a success and were
43. 1:36able to kill bacteria in tests.
44. 1:39>> So, that's what the US team managed to
45. 1:41do, but how big a deal is it? Our health
46. 1:44correspondent James Gallagher assesses
47. 1:46the significance of the research.
48. 1:48>> This is a real step up in the power
49. 1:50[music] of what AI has been able to do
50. 1:52in this kind of field. So, we've already
51. 1:53seen generative artificial intelligence
52. 1:55tools being used to come up with designs
53. 1:57[music] for new antibiotics and for new
54. 1:59vaccines, whereas this is looking at the
55. 2:02whole genetic code, the [music] entire
56. 2:03genome of viruses, generating those in a
57. 2:06computer. And the [music] way that was
58. 2:08done is that scientists were trying to
59. 2:10develop new types of a virus called
60. 2:12phage. And what they do is they infect
61. 2:14bacteria, so that there's no [music]
62. 2:16risk. You don't need to be terrified.
63. 2:18There's no risk to to human beings. They
64. 2:19infect bacteria. And what they did is
65. 2:21they they got the AI to generate the
66. 2:24genetic code, [music] suggested genetic
67. 2:26codes. Then the scientists picked 300
68. 2:28that they thought were the most
69. 2:29promising, [music]
70. 2:30test those in the laboratory. So they
71. 2:31actually then made those viruses as
72. 2:33designed by the AI.
73. 2:35>> [music]
74. 2:35>> And they found that 16 of those were
75. 2:36able to fully function in the
76. 2:38laboratory. So they were infecting
77. 2:40bacteria, they were replicating [music]
78. 2:41inside those bacteria, and that's the
79. 2:44first time this has been done.
80. 2:47>> Well, the breakthrough has been labeled
81. 2:48a very significant turning point in
82. 2:50science that could unlock a new era for
83. 2:53treating disease. But experts have also
84. 2:55warned AI-designed viruses raise urgent
85. 2:58safety and security concerns. Well,
86. 3:00let's start with the benefit. Here's Dr.
87. 3:03Simon Clark, a microbiologist from the
88. 3:05University of Reading.
89. 3:07>> Well, some people have suggested that
90. 3:08you could use these types of viruses
91. 3:10which attack bacteria
92. 3:12to to wipe out to kill bacteria that are
93. 3:15causing infections that are resistant
94. 3:18antibiotics. So it's potentially a novel
95. 3:21strategy for dealing with antibiotic
96. 3:23resistance. There are all sorts of other
97. 3:25problems around that. Um one of which is
98. 3:29that bacteria become resistant to the
99. 3:31viruses as well, not just the
100. 3:33antibiotics. So by generating new
101. 3:36versions of virus, you can get round
102. 3:38that um
103. 3:40resistance. I have to say, when I read
104. 3:42this, I thought, "Yeah, really? You're
105. 3:44You're actually just hanging this
106. 3:46research, this piece of study on that
107. 3:49hook?" Really, they've done this, I
108. 3:51think, because they could. This is an
109. 3:54important proof of principle, and as
110. 3:57AI gets more capable, you might expect
111. 4:01it to be able to
112. 4:03to
113. 4:04tailor or engineer other forms of life.
114. 4:07>> So, what about the risks? Well, concerns
115. 4:10have already been raised that the same
116. 4:12technology could be used maliciously to
117. 4:15create new diseases. Here's Simon Clark
118. 4:17again.
119. 4:18>> Study is not a risk to human health.
120. 4:20It's important to stress that, but it
121. 4:22does show that in theory it is possible
122. 4:25to generate synthetic forms of life,
123. 4:28forms of viruses which don't exist in
124. 4:31nature, and therefore that in itself
125. 4:34could present a risk to human health. We
126. 4:37should say that that that the AI models
127. 4:39that they used here can't do that, but
128. 4:42there's nothing to stop somebody coming
129. 4:44up with a different AI model that
130. 4:46potentially could. This country that
131. 4:48there are all sorts of protections and
132. 4:51and regulations
133. 4:53for for doing studies like this, and
134. 4:55there would be in the United States,
135. 4:57too. But of course,
136. 4:59if a
137. 5:01bad actor was so minded to do something
138. 5:03like that, then they wouldn't really
139. 5:06worry themselves about the normal health
140. 5:09and safety rules that
141. 5:11that control what people like me do.
142. 5:15>> Finally, let's look at the wider impact
143. 5:17of artificial intelligence on health
144. 5:19care. Here again is our health
145. 5:21correspondent, James Gallagher.
146. 5:23>> The more practical tools are not as are
147. 5:26not like this. So, this is kind of like
148. 5:28the the very cutting edge. No one's
149. 5:29entirely sure where this technology
150. 5:31[music] is going to go. The grand vision
151. 5:33of this kind of designing biology in a
152. 5:36computer [music] that could could then
153. 5:38be replicated in the real world, but
154. 5:40isn't actually something that naturally
155. 5:41exists in nature. That could [music]
156. 5:43open up new possibilities for new drug
157. 5:46treatments or these kinds of phage
158. 5:47viruses that infect bacteria. Maybe
159. 5:49these those could be used [music] to
160. 5:51treat bacterial infections. But, it's a
161. 5:53little bit more mundane in terms of how
162. 5:55AI at the moment is being [music] used
163. 5:57in most of healthcare. So, it's things
164. 5:59like transcribing doctor's notes and
165. 6:01providing summaries to make those more
166. 6:03time efficient. There's a lot of work
167. 6:05going on whether artificial intelligence
168. 6:07tools can look
169. 6:08>> [music]
170. 6:08>> at lots of scans, say mammograms as part
171. 6:10of breast cancer screening, and see if
172. 6:12they can be used [music] to make the
173. 6:14whole system more effective and more
174. 6:16efficient. So, it it's a little bit more
175. 6:18mundane than the kinds of very
176. 6:20cutting-edge science that we're talking
177. 6:22about with this study in terms of where
178. 6:24it's being rolled out at the moment.
179. 6:26>> And that's all you need to know about
180. 6:28AI-generated viruses.